Low Endotoxin Recovery - Detection of
Endotoxins in Common Biopharmaceutical
Product Formulations
Dissertation
zur Erlangung des Doktorgrades der Naturwissenschaften
(Dr. rer. nat.)
an der Fakultät für Chemie und Pharmazie
der Universität Regensburg
vorgelegt von
Johannes Reich
aus Regensburg
2016
2
Die Arbeit wurde angeleitet von: Prof. Dr. Hubert Motschmann
Promotionsgesuch eingereicht am: 05.07.2016
Promotionskolloquium am: 04.08.2016
Prüfungsausschuss:
Vorsitz: PD Dr. habil. Axel Dürkop
1. Gutachter: Prof. Dr. Hubert Motschmann
2. Gutachter: Prof. Dr. Dr. h.c. Ulrich Koszinowski
3. Prüfer: Prof. Dr. Joachim Wegener
3
...no matter how many instances of white swans we may have observed, this does not
justify the conclusion that all swans are white.
Karl Popper, 1934, Austrian and British Philosopher
4
5
Table of Content I. Preface .................................................................................................................................7
II. Zusammenfassung ...............................................................................................................9
III. Abstract ............................................................................................................................. 11
IV. Abbreviations .................................................................................................................... 12
1 Fundamentals .................................................................................................................... 13
1.1 Endotoxin (LPS) ............................................................................................................. 13
1.2 Structure and activity of LPS ......................................................................................... 15
1.3 Clinical relevance of LPS ................................................................................................ 16
1.4 Need for control of endotoxins and their detection methods ..................................... 18
1.5 Regulatory aspects ........................................................................................................ 20
1.6 Low Endotoxin Recovery ............................................................................................... 20
1.7 Purpose of the study ..................................................................................................... 21
2 Material and Methods ...................................................................................................... 23
2.1 Reagents ........................................................................................................................ 23
2.2 Endotoxins and bacteria ................................................................................................ 23
2.3 Other materials ............................................................................................................. 24
2.4 Preparation of crude endotoxins extracts .................................................................... 24
2.4.1 Preparation 1 ......................................................................................................... 24
2.4.2 Preparation 2 ......................................................................................................... 24
2.5 Sample handling ............................................................................................................ 25
2.5.1 Addition of endotoxin to a sample ........................................................................ 25
2.5.2 Sample preparation before endotoxin detection ................................................. 25
2.5.3 Incubation of endotoxin in a sample ..................................................................... 25
2.5.4 Preparation of endotoxin recovery kinetics .......................................................... 25
2.5.5 Endotoxin recovery and masking controls ............................................................ 26
2.5.6 Sample preparation for demasking of endotoxin ................................................. 26
2.6 Limulus-based endotoxin detection assays .................................................................. 27
2.6.1 Recombinant Factor C assays (rFC) ....................................................................... 27
2.6.2 Limulus Amebocyte Lysate assay (LAL) ................................................................. 27
2.7 Silver stained SDS-PAGE ................................................................................................ 27
2.8 Calculations and plots ................................................................................................... 28
3 Results and Discussion ...................................................................................................... 29
3.1 Masking of endotoxin in surfactant samples: Effects on Limulus-based detection
systems ...................................................................................................................................... 29
6
3.1.1 Introduction ........................................................................................................... 29
3.1.2 Results ................................................................................................................... 31
3.1.3 Discussion .............................................................................................................. 37
3.2 Endotoxin Masking: A kinetically controlled reaction mechanism ............................... 41
3.2.1 Introduction ........................................................................................................... 41
3.2.2 Results ................................................................................................................... 42
3.2.3 Discussion .............................................................................................................. 49
3.3 Heterogeneity of endotoxins and their detectability in common biopharmaceutical
formulations .............................................................................................................................. 54
3.3.1 Introduction ........................................................................................................... 54
3.3.2 Results ................................................................................................................... 56
3.3.3 Discussion .............................................................................................................. 63
3.4 Demasking of Endotoxin ................................................................................................ 68
3.4.1 Introduction ........................................................................................................... 68
3.4.2 Results ................................................................................................................... 71
3.4.3 Discussion .............................................................................................................. 79
4 Conclusions ........................................................................................................................ 86
4.1 Endotoxin demasking – a technical solution ................................................................. 86
4.2 Perspectives of sample treatments in BET .................................................................... 87
4.3 Need for structural analysis of endotoxins in complex sample matrices ..................... 88
4.4 Clinical relevance of masked endotoxin ........................................................................ 89
4.5 Key findings ................................................................................................................... 91
5 Appendix ............................................................................................................................ 93
5.1 List of figures ................................................................................................................. 93
5.2 List of tables .................................................................................................................. 95
5.3 References ..................................................................................................................... 96
5.4 List of publications ....................................................................................................... 107
5.4.1 Selected poster presentations............................................................................. 107
5.4.2 Oral presentations ............................................................................................... 107
5.4.3 Published articles................................................................................................. 108
5.4.4 Intellectual Property ............................................................................................ 108
5.4.5 Declaration .......................................................................................................... 109
5.5 Curriculum Vitae .......................................................................................................... 110
5.6 Eidesstattliche Erklärung ............................................................................................. 111
6 Annex ............................................................................................................................... 113
7
I. Preface This work was carried out between October 2012 and June 2016 at the Hyglos GmbH, Research
and Development Department, Bernried, Germany in collaboration with the University of
Regensburg, Institute of Physical and Theoretical Chemistry, Regensburg, Germany under the
supervision of Prof. Dr. Hubert Motschmann.
This work would not have been possible without the help and guidance of the following people:
First of all, I would like to express my sincere gratitude to Prof. Dr. Hubert Motschmann from
University of Regensburg, who enabled this interdisciplinary project and supervised my thesis.
Thank you for guiding my work in an ambitious way and having faith in my work and decisions.
Parts of this work could not have been realized without the close collaboration with Hyglos. I am
likewise grateful to Dr. Wolfgang Mutter, who supported the industrial application and mentored
the attendance in international endotoxin working groups. Beyond that I want thank Hyglos for
the financial support.
Furthermore, my special thanks go to Dr. Holger Grallert from Hyglos for introducing me to the
fascinating, comprehensive science about endotoxins and for the numberless, long and very
valuable discussions.
Likewise, I would like to thank the whole research and development group at Hyglos. Especially,
I would like to thank Martin Vogel, Katharina Heinzelmann, Anja Schlicht, Rebecca Jörg and Petra
Schneider, who helped me with some experiments during this work. Many thanks also go to the
trainees Frederik Kappelt and Franziska Schmitz for their help with some experiments.
Furthermore, I want to thank Dr. Frederik Baumkötter for critical reading parts of this manuscript.
I also would like to express my gratitude to Dr. Bernd Buchberger from Microcoat Biotechnologie
GmbH, Bernried, Germany for path-breaking and thought-provoking impulse on experimental
setups.
In particular, I would like to thank Prof. Dr. Jack Levin from University of California, San Francisco,
USA, Prof. Dr. Ulrich Zähringer from Forschungszentrum Borstel, Borstel, Germany, Dr. Hiroshi
Tamura from Juntendo University School of Medicine, Tokyo, Japan and Prof. Dr. Ulrich
Koszinowski from Ludwig Maximilian Universität, München, Germany for sharing their
outstanding experience about endotoxins and giving me unaffordable advice in all matters.
I also would like to thank Dr. Steven Zeman and Dr. Franz-Josef Zimmer from Gruenecker Patent-
und Rechtsanwälte PartG mbB, München, Germany in all questions about the development and
application of a patent.
Furthermore, I would like to thank Sylvia Becker and Dr. Georg Rössling from PDA Europe
gemeinnützige GmbH, Berlin, Germany for the warm welcome to the scientific committee of the
European pharmaceutical microbiology conference. Moreover, my sincere thanks go to the PDA
for admitting me to the low endotoxin recovery task force, and special thanks go to the chairman
8
Dr. Friedrich von Wintzingerode, Roche Diagnostics GmbH, Penzberg, Germany and Dr. Dayue
Chen, Eli Lilly and Company, Indianapolis, USA.
Likewise, many thank go to Axel Schroeder from Concept Heidelberg GmbH, Heidelberg, Germany
for inviting me to diverse trainings, meetings and webinars and the subsequent discussions about
low endotoxin recovery.
I also would like to thank all collaborating partners in pharmaceutical industry for cooperative
working partnerships and for giving trust in our advanced technologies.
Moreover, many thanks go to Pierre Lang from F. Hoffmann-La Roche Ltd, Basel, Switzerland
(retired since 2015) and Kevin Williams from General Electric, Boulder, USA for great discussions
about low endotoxin recovery and innate immune response modulating impurities.
Furthermore, I would like to thank Dr. Patricia Hughes from CDER, Food and Drug Administration,
Silver Spring, USA for giving me insights into regulatory requirements for biological license
applications of biopharmaceutical drug products during several round table and panel discussions
and I highly appreciate the scientifically based considerations of the low endotoxin recovery
phenomenon.
Last but not least, I would like to express my deepest thanks to my family and very special thanks
go to Elisabeth - thank you infinitely for everything - especially for giving me strength and
incredible encouragement in hard times.
Johannes Reich
9
II. Zusammenfassung Bakterien gehören zu den ersten und ältesten Lebewesen auf der Erde und aufgrund ihrer
Allgegenwärtigkeit ist der Mensch permanent in direktem Kontakt mit ihnen. In dieser Arbeit wird
einem Abbauprodukt der Gram-negativen Bakterien, den sogenannten Endotoxinen, spezielle
Aufmerksamkeit gewidmet. Diese sind der Hauptbestandteil der äußeren Membran von Gram-
negativen Bakterienzellen und werden während des Wachstums oder beim Absterben und der
Lyse freigesetzt. Endotoxine sind für den Menschen generell nicht gesundheitsschädlich, solange
sie nicht in den Blutkreislauf eingetragen werden. In diesem Fall können sie jedoch zu
schwerwiegenden pathogenen Wirkungen führen (z.B. Sepsis). Ein potentielles Risiko der
Übertragung in den Blutkreislauf ist beispielsweise bei der Injektion von Medikamenten gegeben.
Um dieses Risiko zu minimieren, sind Hersteller von injizierbaren Medikamenten (Parenteralia)
verpflichtet, diese streng auf Endotoxin-Kontaminationen zu überprüfen. Hierzu wird
üblicherweise der „Limulus Amöbozyten Lysat“ Test verwendet. Dieses Testsystem basiert auf
enzymatischen Reaktionen, die der Blutgerinnungskaskade von Pfeilschwanzkrebsen
entstammen. Aufgrund der Sensitivität und Anwenderfreundlichkeit stellt diese Methode seit
Jahren den „Goldstandard“ zur Bestimmung von Endotoxinen dar. Seit neuestem werden jedoch
Unstimmigkeiten während der Testung von biopharmazeutischen Medikamenten beobachtet. In
bestimmen Medikamenten konnten die Positivkontrollen nicht innerhalb bestimmter
Akzeptanzkriterien detektiert werden. Dieser Effekt wird als „Low Endotoxin Recovery (LER)“
bezeichnet und führt zu Bedenken auf Anwender- und Behördenseite, ob diese Testmethode
weiterhin zuverlässig und anwendbar ist.
In dieser Arbeit wurde die Detektierbarkeit von Endotoxin in biopharmazeutischen
Formulierungen mittels Limulus-basierten Testmethoden untersucht, um die Ursachen des LER-
Phänomens zu verstehen und die gegenwärtigen Testmethoden zu optimieren. Die präsentierten
Daten machen deutlich, dass allgemein verwendete Formulierungsbestandteile der
Medikamente den LER-Effekt verursachen können. Als Hauptursache wurde die simultane
Anwesenheit von Tensiden und Chelatoren identifiziert. Des Weiteren wird gezeigt, dass LER ein
zeitabhängiges Phänomen ist und die Reaktionsgeschwindigkeit hauptsächlich von der
Konzentration des Chelators in der Probenmatrix abhängt. Darüber hinaus wurden verschiedene
Endotoxine untersucht, die aufgrund ihrer strukturellen Heterogenität unterschiedliche
Empfindlichkeiten hinsichtlich der Detektierbarkeit aufweisen. Zusammengenommen deuten die
Daten darauf hin, dass sich das Endotoxin in einem zweistufigen Prozess in Tensidmizellen der
Probenmatrix einlagert und somit für die Detektion maskiert ist. Mittels dieser Arbeitshypothese
wurde eine „Toolbox“, mit verschiedenen amphiphilen und chaotropen Reagenzien zur
Demaskierung des Endotoxins entwickelt. Schließlich wird gezeigt, dass durch eine gezielte
10
Probenvorbereitung der LER-Effekt aufgehoben und das Endotoxin mit herkömmlichen Limulus-
basierten Testmethoden wieder detektiert werden kann.
11
III. Abstract Bacteria are one of the first and oldest living organisms on earth and due to their ubiquity,
humans are permanently in close contact to them. This work focuses on special breakdown
products of Gram-negative bacteria so-called endotoxins. Endotoxins are the major component
of the outer membrane of Gram-negative bacteria. They are released during growth or death and
lysis of the bacterial cell. Endotoxins are usually not hazardous to man, as long as they do not
enter the circulating blood system. However, if for example endotoxins enter the blood system
they can lead to severe pathogenic effects like sepsis. One prominent risk is the accidental
injection of contaminated drug products. In order to reduce this risk, manufacturers of parenteral
drug products are forced to meticulously control their products before they are released to the
market. To this end, Limulus Amebocyte Lysate (LAL) assays have been the gold standard for
detection of bacterial endotoxins since years. These assays are based on enzymatic reactions
derived from the blood coagulation cascade in horseshoe crabs. Most quality control
departments in pharmaceutical industry have established such methods to release their drug
products, because these assays enable fast and sensitive results. However, in the recent past
inconsistencies during testing of biopharmaceutical drug products have been observed. In certain
drug products, positive controls of endotoxin were not detectable within given acceptance
criteria. This effect is called Low Endotoxin Recovery (LER) and users as well as authorities are
concerned about the reliability of the existing test procedures.
In this work, the detectability of endotoxin in biopharmaceutical drug product matrices using
Limulus-based detection systems was analyzed to understand the phenomenon of LER and to
optimize existing test procedures. The results show that common formulation components of
biopharmaceutical drug products are capable to induce LER. The minimum prerequisite is the
simultaneous presence of surfactants and complex forming agents. It is demonstrated that the
occurrence of LER is time-dependent and that the reaction rate of LER is substantially depending
on the concentration of the complex forming agents in the sample matrix. Moreover, endotoxins
from different sources were studied, because their structural heterogeneity may lead to different
masking susceptibilities. Together, these results indicate that LER is caused by masking of
endotoxin leading to an alteration of its supramolecular structure, in which endotoxins are
embedded in surfactant micelles. The further elucidation of a two-step LER reaction mechanism
served as a basis for the development of a toolbox including amphiphilic and chaotropic reagents,
which enables the demasking of endotoxin. In conclusion, the dedicated sample treatment using
such reagents allows the detection of LER-affected endotoxin by Limulus-based detection
methods.
12
IV. Abbreviations
API Active Pharmaceutical Ingredient
Ara4N 4-Amino-4-Deoxyarabinose
B BSA
B.cepacia Burkholderia cepacia
BD BSA + Dodecanol
BET Bacterial Endotoxin Testing
BSA Bovine Serum Albumin
C Calcium Dichloride
Ci Citrate
Ca2+ Calcium2+
CBSD Calcium dichloride+ BSA+ SDS + Dodecanol
CMC Critical Micelle Concentration
CSE Control Standard Endotoxin
D Dodecanol
E.cloacae Enterobacter cloacae
E.coli Escherichia coli
EDTA Ethylenediaminetetraacetic Acid
EU Endotoxin Unit
FDA Food and Drug Administration
HLB Hydrophilic Hydrophobic Balance
LAL Limulus Amebocyte Lysate
LB Lysogeny Broth
LER Low Endotoxin Recovery
LPS Lipopolysaccharides
MAT Monocyte Activation Test
NOE Naturally Occurring Endotoxin
OMV Outer Membrane Vesicle
P.aeruginosa Pseudomonas aeruginosa
PBS Phosphate-Buffered Saline
P.mirabilis Proteus mirabilis
PPC Positive Product Control
R.pickettii Ralstonia pickettii
rFC Recombinant Factor C
rpm Round per Minute
RPT Rabbit Pyrogen Test
RSE Reference Standard Endotoxin
RT Room Temperature
S SDS
S.maltophilia Stenotrophomonas maltophilia
S.marcescens Serratia marcescens
SDS Sodium Dodecyl Sulfate
SDS-PAGE Sodium Dodecyl Sulfate-Polyacrylamide Gel Electrophoresis
TLR Toll Like Receptor
TNF Tumor Necrosis Factor
TRIS 2-Amino-2-(Hydroxymethyl)-1,3-Propanediol
13
1 Fundamentals Bacteria are one of the first and oldest living organisms on earth and their diversity is practically
unlimited. Since modern analytical techniques enable the determination of genetic codes in
bacteria, they can be categorized by their molecular phylogeny[1]. However, for historical and
practical reasons bacteria are also classified by structural differences in their cell walls. Based on
chemical and physical properties of the cell walls, bacteria can be differentiated between Gram-
positive and Gram-negative. This dates back to times in which bacteria were stained (e.g. methyl
violet) to enhance the visibility of bacteria [2].
1.1 Endotoxin (LPS) Endotoxins are a unique group of molecules, which occur naturally in the cell wall of Gram-
negative bacteria[3]. They are derived for example from various bacteria like Neisseria
meningitides, Vibrio cholera or Escherichia coli (Figure 1A). If administrated into the blood stream
of mammalians, bacteria and their toxic breakdown products can cause severe pathogenic effects
including fever and septic shock[4]. Due to their fever inducing capability, endotoxins are also
classified as pyrogens.
Figure 1 Gram-negative Bacteria
A) Scanning electron micrograph of Escherichia coli, grown in culture and adhered to a cover slip. (Credit: National Institute of Health, National Institute of Allergy and Infectious Diseases, Image Library). B) Schematic representation of the Gram-negative membrane. The outer membrane possesses an asymmetric bilayer, in which LPS covers mostly the surface and phospholipids are located in the inner leaflet as counter molecules. In addition, a variety of further components like proteins, lipoproteins, peptidoglycans are present in the cell envelope of Gram-negative bacteria. (Source: [4]) C) Chemical structure of LPS from E. coli O111:B4 LPS. The particular regions of LPS (lipid A, core region and O-antigen) and the variability of sugar units are indicated. (Source: [5])
14
The end of the 19th century is of particular importance in the context of endotoxins. Dramatic
cholera outbreaks threatened large harbor cities including Hamburg, Germany and St. Petersburg,
Russia and have led to a large number of death. During that time, systematic research began and
is still ongoing to better understand the hazardous effects of Gram-negative bacteria and
endotoxins (Figure 2).[6]
Figure 2 The history of endotoxin research
Research on endotoxins started in the end of 19th century with the description of the deleterious and
beneficial bioactivities of endotoxins. A further phase comprised efforts undertaking their biochemical
and immunochemical characterization, while the immunological properties of this molecule were
mainly defined in the mid-20th century. (Source:[7])
Around 1890, Richard Pfeiffer, a co-worker of Robert Koch, stated:
“In ganz jungen, aerob gezüchteten Choleraculturen ist ein specifischer Giftstoff
enthalten, welcher ausserordentlich intensive toxische Effecte entfaltet. Dieses
primäre Choleragift steht in sehr enger Zusammengehörigkeit zu den
Bacterienleibern und ist vielleicht ein integrirender Bestandtheil derselben. Durch
Chloroform, Thymol und durch Trocknen können die Choleravibrionen abgetödtet
werden, ohne dass dieser Giftstoff anscheinend verändert wird.1”[8]
1 In very young aerobically grown cholera cultures, a specific toxic substance is contained which exerts extraordinarily intense toxic effects. This primary cholera toxin is closely attached to, and probably an integral part of, the bacterial body. By chloroform, thymol, or by drying, the cholera vibrios can be killed without any detectable change of the toxin[6].
15
This was the genesis of the concept of endotoxins. Within the mid of the 20th century, it was
figured out that endotoxins are located at the surface of Gram-negative bacteria (Figure 1B) and
are liberated when bacteria multiply, die and lyse[9]. During the development of techniques for
extracting and preparing endotoxins, lipopolysaccharides (LPS) (Figure 1C) were identified in
bearing the toxic principle of endotoxin[3]. To this end, the two terms endotoxin and LPS are used
for the same molecule and thus represent synonyms. However, the term endotoxin reflects its
biological activity and the term LPS its chemical composition[10]. Moreover, it has been shown
that LPS are the dominating constitutes of the outer membrane of Gram-negative bacteria and
LPS play an important role in maintaining the integrity of the membrane architecture and is
therefore an essential component for bacterial viability. Noteworthy, LPS are accompanied by
certain proteins and lipids, but LPS covers three-quarters of the bacterial surface and one
bacterial cell contains approximately 3.5 x 106 LPS molecules.[9]
1.2 Structure and activity of LPS LPS are a broad and complex group of molecules, which possess a common general
architecture[11]. The molecules can be divided in three parts: O-antigen, core region and lipid A
(Figure 3)[12]. The latter is based on a phosphorylated diglucosamine which is esterified with fatty
acids (e.g. caproic acid (C6), lauric acid (C12), myristic acid (C14), palmitic acid (C16), stearic acid
(C18)) and anchors the molecule in the outer membrane of Gram-negative bacteria and is
covalently substituted by a saccharide portion. The core region comprises an oligosaccharide
containing up to fifteen monosaccharides (e.g. diverse heptoses, glucose, galactose and
mannooctulosonic acid) to which a polysaccharide portion of repeating units, the O-specific chain
including glucose, galactose, rhamnose and mannose, is linked.[10] Noteworthy, also LPS mutants
lacking the O-antigen have been isolated. These LPS forms are donated as rough LPS, whereas LPS
containing O-antigen are called smooth LPS[13]. However, it has been shown that the lipid A part
represents the “endotoxic principle” of LPS, and in contrast to the role of the polysaccharides
(core region and O-antigen), alterations of the lipid A moiety were found to influence the
bioactivity dramatically[14], [15]. Full endotoxic activity is expressed by a molecule containing
two hexosamine residues, two phosphoryl groups, and six fatty acids including 3-acyloxyacyl
groups with a defined chain length and at a distinct location[9]. As a consequence, not all LPS are
toxic, just as not all bacteria are pathogenic[3]. Moreover, due to the hydrophilic (O-antigen and
core region) and hydrophobic (lipid A) regions, LPS tend to aggregate in aqueous solutions.
Depending on the molecular structure and environmental conditions diverse supramolecular
structures either non-lamellar inverted (cubic Q or hexagonal HII) or lamellar can be formed[16].
It has been shown in several studies that the aggregation state of LPS affects its biological activity
as well as its detectability.[14], [16]–[19]
16
Figure 3 Schematic structure of lipopolysaccharides
With regard to the biological nature, LPS can be divided into three functional subunits: O-antigen, core region and lipid A. The latter is the toxic fragment of the molecule. With regard to the chemical structure, LPS is an amphiphilic molecule. The fatty acids within the lipid A are hydrophobic and the polysaccharides in the core region and O-antigen are hydrophilic. In addition, LPS is electrically charged due to substituents (e.g. phosphates) in the core region and on the diglucosamine of lipid A.
1.3 Clinical relevance of LPS Years ago, clinicians have recognized that humans are especially sensitive towards endotoxin. In
some cases, intravenous infusions containing bacterial contaminants have led to death and
severe pathogenic response in patients.[20]–[22] Thereby, LPS are able to induce a variety of
biological effects in-vivo (Figure 4). To fight against pathogenicity, bacteria and its LPS are the
primary target for interaction with antibacterial drugs and components of the immune system of
the host.[17]
Figure 4 Schematic representation of the activation mechanisms induced by LPS
LPS is released from the bacterial outer membrane by the attack of immune components (drugs, proteins) or simply by cell division. It may interact with serum and membrane proteins, which subsequently leads to an activation of macrophages, which secret mediators such as tumor necrosis factors alpha and interleukins. (Source: [17])
17
In order to understand pathogenicity, an important finding has been the identification of the
plasma membrane protein Toll-like receptor 4 (TLR4) as the lipid A signaling receptor of animal
cells. Activation of TLR4 by lipid A triggers the biosynthesis of diverse mediators of inflammation
including tumor necrosis factor (TNF) or interleukins, ultimately resulting in multiple organ failure,
septic shock in the case of systemically overproduction.[23] Generally, TLR4 belongs to a family
of innate immunity receptors (Figure 5) and besides endotoxins, there are also other pathogens,
which are not limited to Gram-negative bacteria (Lipoteichoic acids, peptidoglycans, proteins etc.)
and stimulate the innate immune system resulting in pyrogenic reactions. However, endotoxins
are considered to be an outstanding alarm marker due to their relatively high pyrogenicity.
Endotoxins are active in the picogram range per kilogram bodyweight. Therefore, a little amount
of endotoxin can generate a very strong host response.[24] In contrast, peptidoglycans are 50,000
times less pyrogenic than endotoxins[25].
Figure 5 The Toll-Like-Receptor family
Toll-like receptors (TLRs) recognize a variety of pathogen-associated molecular patterns. Recognition of LPS by TLR4 is aided by accessory proteins (CD14 and MD-2). TLR2 recognizes a broad range of structurally unrelated ligands and functions in combination with several (but not all) other TLRs, including TLR1 and TLR6. TLR3 is involved in recognition of double-stranded RNA. TLR5 is specific for bacterial flagellin, whereas TLR9 is a receptor for unmethylated CpG sequences in DNA. (Source [26])
For further comprehension of the pyrogenicity of endotoxins, dose-febrile response curves for
endotoxins have been studied and it was found that man, cat, horse and rabbits have
approximately the same threshold to pyrogen simulation by endotoxins. However, larger doses
are more pyrogenic and more toxic for man than for rabbit, due to much steeper dose-febrile
response curves for man. Dogs and chimpanzee were notably less susceptible to the pyrogenicity
of endotoxin than the other species.[27] Moreover, depending on the source of endotoxin
different thresholds were found to initiate pyrogenicity. For example approximately 0.1 to 1.0
nanogram per kilogram of bodyweight of endotoxin from E.coli is needed, whereas 50 nanogram
per kilogram of bodyweight of endotoxin from P.aeruginosa are required for pyrogenic response
in man[28]. This already gives an indication about the heterogeneity and complexity of
18
endotoxins. In order to enable a comparison of biological activities, endotoxin units (EU) were
introduced, based on the approximated threshold of pyrogenic activity of E.coli. 1 EU reflects the
biologic activity of 0.1 nanogram of purified endotoxin from E.coli in man.[24]
1.4 Need for control of endotoxins and their detection methods The occurrence of Gram-negative organisms in virtually every environment on earth makes LPS
one of the most prevalent complex organic molecules occurring in nature. Gram-negative
bacteria have been isolated wherever man has gone: in soil, fresh and salt water, frigid oceans
and hot springs. Minimal growth requirements of Gram-negative bacteria allow their growth in
the cleanest of water.[24] Therefore, the ubiquity of endotoxins requires routine screening of all
fluids and medications prepared for parenteral therapy.[27] Although the pyrogenic dose
response curve in man is much steeper than it is in rabbits, the minimum pyrogenic dose on a
weight basis in rabbits is in a passable range compared to man[28]. Hence, the Rabbit Pyrogen
Test (RPT) was introduced as an in-vivo test method for the detection of fever-causing (pyrogenic)
contaminations in pharmaceutical products, and has already been manifested in various
pharmacopoeias and guidelines around the world in the 1940th[29]. Over the years, alternative
in-vitro detection methods were discovered and established. One method of pyrogen detection
relies on mimicking the human fever reaction, which can be found in the Monocyte Activation
Test (MAT). It employs the cytokine response of blood monocytes for the detection of
microbiological contaminants.[30] However, handling of appropriate blood or cell lines for
running the assay and regulatory issues prolong the universal application of this assay.
Another and more prominent method was discovered by Bang and Levin in the 1960th. They
utilized the defense system of an animal with over 450 million years of experience, fighting
against microbial attacks. It has been demonstrated that bacterial endotoxins rapidly induce
clotting of the blood of horseshoe crabs[31]. The amebocytes in horseshoe crabs´ hemolymph
contain a coagulation system, which is activated by minute amounts of endotoxin.[32] The
principle of this test method is based on an enzymatic reaction cascade. In the presence of LPS,
an LPS-sensitive serine protease zymogen Factor C, is autocatalytically activated. The active
Factor C* then activates zymogen Factor B to Factor B*, which subsequently activates proclotting
enzyme to clotting enzyme. The resulting clotting enzyme converts soluble coagulogen, an
invertebrate fibrinogen-like substance, to an insoluble coagulin gel. (Figure 6B)[33]. This cascade
system found in the hemocytes allows for an extremely high sensitivity of the lysate to picogram
quantities of endotoxins. [34] For production of the so called Limulus Amebocyte Lysate (LAL),
horseshoe crabs (Figure 6A) are collected when they migrate to shallow coastal waters for
reproduction. Once collected, the lively crabs are placed in restraining racks. Sterile needles are
19
inserted through the muscular hinge between the cephalothorax and abdominal region[24], [32]
and up to 150 mL blood of one horseshoe crab can be obtained. If the crabs are handled with
care, they normally survive this procedure. Depending on country-specific regulations they are
returned into the ocean or are further processed (e.g. fishing bait). However, after collection the
blood is centrifuged and the harvested amebocytes are washed. The cells are lysed by addition of
distilled water and cellular debris is removed by centrifugation. Finally, the lysate is decanted and
can be used for testing.[35] Comparative studies between LAL tests and the RPT on various
samples resulted in good agreement and the results achieved by LAL detection methods were
more sensitive, when samples were properly handled[36], [37].
Figure 6 Horseshoe crabs and their endotoxin specific reaction cascade
A) Horseshoe crabs at Pickering beach, Delaware, USA. Horseshoe crabs are found along the northeast coasts in America and southeast costs in Asia. For reproduction, adult crabs travel during late spring and early summer from deep ocean water to coastal water and females deposit eggs on the beaches. B) Tentative reaction mechanism for the coagulation cascade of the Limulus amebocyte lysate with endotoxin.
Practical experience and technological progress led from simple test techniques, based on the
occurrence of gel formation by the reaction of the lysate with endotoxin; to photometric
techniques, which are based on the change in lysate turbidity during gel formation; and
chromogenic techniques, which are based on the development of color after cleavage of a
synthetic peptide-chromogen complex[38]–[40]. Moreover, growing demand and limited
resources of horseshoe crabs are leading to the use of recombinant sources of Limulus-based
enzymes. These novel techniques use, recombinantly produced Factor C (rFC), the first enzyme
of the Limulus coagulation cascade, and a fluorogenic substrate is generating the signal[41], [42].
In the present work, mainly recombinant test methods were used. Today, these Limulus-based
methods including LAL and rFC are recognized as the most sensitive in-vitro assays available for
20
bacterial endotoxin testing (BET). These methods are more economical and require a smaller
volume of sample for testing than does the RPT and MAT. In addition, a large number of tests can
be performed by one individual in a single day.
1.5 Regulatory aspects Endotoxin is only a concern for man, when it comes into contact with the circulatory blood
system. One relevant mechanism for such contact involves medically invasive techniques
including injection or infusion of parenteral solutions[24]. Therefore, pharmaceutical regulatory
agencies around the world are asking for BET in parenteral drug products. For example, the
European Directorate for the Quality of Medicines (EDQM) states that
“bacterial endotoxins are the most common cause of pyrogenicity in pharmaceutical
products. Any preparation administered parenterally should be sterile and comply
with the test for bacterial endotoxins. Substances to be used in parenteral
preparations must comply with the BET, whatever their origin…”[43]
In consequence, manufacturers of parenteral drug products are obliged to perform BET of in-
process samples and finished products. Fortunately, the occurrence of contaminations in
parenteral drugs, devices, infusions and transfusion solutions has been relatively rare since the
introduction of BET.[24] In order to maintain such a high level, the meticulousness in quality
control of pharmaceutical products has to retained and continuously improved.
1.6 Low Endotoxin Recovery With regard to certain pharmaceutical drug products some inconsistencies were recognized
during BET. In 2013, Chen and Vinther from Genentech reported the phenomenon of Low
Endotoxin Recovery (LER)[44]. During the establishment of diverse test procedures, to simulate
potential contamination events, known amounts of endotoxin were inadequate detectable.
Defined amounts of endotoxin were spiked to undiluted drug products and these samples
containing endotoxins were incubated for a certain period of time. After incubation, the detection
of spiked endotoxin contents resulted in low endotoxin recovery. Interestingly, this phenomenon
was mainly observed in biopharmaceutical drug products, in which the Active Pharmaceutical
Ingredients (API) are proteins like monoclonal antibodies. For stability reasons, such products are
commonly formulated using excipients like phosphate and citrate buffer systems and
polysorbates (Table 1). First investigations of affected samples have indicated that excipients of
the drug products provoke the phenomenon of LER, but non-harmonized test procedures in
industries leading to confounding results. The observation of this phenomenon is meanwhile
frequently discussed in many forums, as it may result in an underestimation of hazardous
endotoxin contents in injectable drug products. The Food and Drug Administration (FDA) pointed
21
out that endotoxin might be present in high amounts in a certain drug product and current assays
are not detecting it or only detecting “acceptable” levels. Depending on the drug product dose
and the potential amount of endotoxin contamination a pyrogenic reaction could occur.[45] In
conclusion, to avoid the underestimation of a potential endotoxin contamination, the driving
forces of the LER phenomenon have to be understood and current test procedures have to be
optimized to ensure entire endotoxin detection in biopharmaceutical drug products.
Table 1 Common formulations of biopharmaceutical drug products
Monoclonal antibodies are prominent active pharmaceutical ingredients (API) in commercial biopharmaceutical drug products. For formulation excipients like phosphate, citrate, sodium chloride and polysorbates are used. (Source: http://www.rxlist.com)
Commercial Drug Product:
Active Ingredient:
Formulation components:
Actemra ® Tocilizumab Phosphate, Sucrose, Polysorbate 80
Avastin ® Bevacizumab Phosphate, Trehalose, Polysorbate 20
Erbitux ® Cetuximab Phosphate, Sodium chloride
Humira ® Adalimumab Phosphate, Citrate, Mannitol, Sodium chloride, Polysorbate 80
Lucentis ® Ranibizumab Histidine, Polysorbate 20
Mabthera ® Rituximab Citrate, Sodium chloride, Polysorbate 80
Remicade ® Infliximab Phosphate, Sucrose, Polysorbate 80
Simponi ® Golimumab Histidine, Sorbitol, Polysorbate 80
Soliris ® Eculizumab Phosphate, Sodium chloride, Polysorbate 80
Synagis ® Palivizumab Histidine, Glycine, Mannitol
Tysabri ® Natalizumab Phosphate, Sodium chloride, Polysorbate 80
Xolair ® Omalizumab Histidine, Sucrose, Polysorbate 20
1.7 Purpose of the study Aim of this work is to improve existing test procedures and to detect endotoxin out of samples,
which are affected by the LER phenomenon. First of all, the phenomenon has to be analyzed
according to the questioning observations, which were made in pharmaceutical industry.
Therefore, endotoxin recovery in common formulation components of drug products including
citrate and phosphate buffer systems as well as surfactants like polysorbates has to be
investigated. Due to the temporally delayed occurrence of the LER-effect, reaction kinetics has to
be examined in order to identify the time limiting parameters of LER and to establish a model
system, which enables the simulation of such kinetics. In addition, as endotoxins represent a
heterogeneous group of LPS, it has to be examined which influences this variability has on the
LER-effect and if it is depending on a certain species of endotoxin. Therefore, endotoxins from
different sources have to be tested with regard to their detectability under LER conditions. After
analysis of the driving forces, an approach to render the endotoxin detectable has to be figured
22
out. Due to the regulatory scope in pharmaceutical industry, existing test procedures have to be
maintained, but advanced by sample treatments prior to the actual measurement. Thus, purpose
of this work is to discover “demasking” agents for sample pre-treatment to detect endotoxin out
of LER-affected samples using Limulus-based test methods and in turn to reduce the risk of wrong-
negative test results during quality control of drug products in pharmaceutical industries.
23
2 Material and Methods
2.1 Reagents Polysorbate 20, polysorbate 80, octoxynol 9, ethanol, 1-octanol, 1-decanol, 1-dodecanol, 1-
tetradecanol, 1-hexadecanol, 1-octadecanol, sodium chloride, sodium azide, citric acid, trisodium
citrate, phosphoric acid, sodium dihydrogenphosphate, potasium dihydrogenphosphate,
disodium hydrogen phosphate-heptahydrate, ethylenediaminetetraacetic acid (EDTA), 2-amino-
2-(hydroxymethyl)-1,3-propanediol (tris), 2-mercaptoethanol, isopropanol, D(+)-glucose, sodium
chloride, calcium dichloride and magnesium dichloride were obtained from Sigma-Aldrich Chemie
GmbH, Steinheim, Germany. Ammonium hydroxide, formaldehyde and D(+)-trehalose-dihydrate
were obtained from AppliChem GmbH, Darmstadt, Germany. Acetic acid, glycerol, periodic acid,
sodium hydroxide, silver nitrate, sodium dodecylsulfate (SDS) and yeast extract (powdered) were
obtained from Carl Roth GmbH & Co.KG, Karlsruhe, Germany. Bromophenol blue sodium salt was
obtained from Merck Chemicals GmbH, Darmstadt, Germany. Bovine Serum Albumin (BSA) and
20x Tris-tricine/SDS electrophoresis buffer were obtained from Serva Electrophoresis GmbH,
Heidelberg, Germany. A bovine polyclonal immunoglobulin G (PAK) and a mouse monoclonal
antibody (MAK33) were obtained from Roche Diagnostics Deutschland GmbH, Mannheim,
Germany. Tryptone Bacto TM was obtained from Becton Dickinson GmbH, Heidelberg, Germany.
Prior to the experiments, all relevant materials have been tested on endotoxin contents and were
proven to contain less than 0.05 EU/mL.
2.2 Endotoxins and bacteria Endotoxin from E.coli O55:B5 (gel-filtrate), P.aeruginosa and S.enterica were obtained from
Sigma-Aldrich Chemie, Steinheim, Germany. Phenol-extracted clinical isolate endotoxins from
E.coli, K.pneumonia, M.morganii, Y.enterocolitica, N.meningitis, P.mirabilis and S.marcescens
were a kind gift from Dr. Andreas Wieser, Universitätsklinik München (LMU), München, Germany.
Endotoxin from E.coli K12 was obtained from InvivoGen, Toulouse, France. Freeze dried bacteria
from E.coli O55:B5 (DSM 4779), E.cloacae (DSM 30054) and P.aeruginosa (DSM 500 71) were
obtained from Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures,
Braunschweig, Germany. Freeze dried bacteria from E.coli O113 (Ecor 30) was obtained from
Escherichia coli Reference Collection, East Lansing, USA. Freeze dried bacteria from B.cepacia
(2008 B02-12.20.164) and S. maltophilia (DSMZ 50 170) were obtained from Robert Koch-
Institute, Wernigerode, Germany. Bacteria from R.pickettii (isolate) were a kind gift from Hyglos
GmbH, Bernried, Germany.
24
2.3 Other materials 20% gradient tris-tricin gel was obtained from Anamed Elektrophorese GmbH, Rodau, Germany.
0.2 µm Acrodisc 25 mm Syringe Filters were obtained from Pall GmbH, Dreieich, Germany. Kinetic
chromogenic Limulus Amebocyte Lysate test was obtained from Lonza Inc., Walkersville, USA.
Depyrogenated water, depyrogenated borosilicate glass tubes and recombinant Factor C tests,
EndoZyme® and EndoLISA® were obtained from Hyglos GmbH, Bernried, Germany.
2.4 Preparation of crude endotoxins extracts
2.4.1 Preparation 1
For growth of bacteria, 5 mL lysogeny broth (LB) media (10 g/L sodium chloride, 5 g/L yeast extract
and 10 g/L tryptone) were inoculated with desired bacterial strain, followed by incubation at 37
°C in a shaking incubator (Platform shaker: Innova 2300, New Brunswick Scientific Co, Enfield,
USA; Incubator: Wärmeschrank für Plattformschüttler, Mytron Bio- und Solartechnik GmbH,
Heilbad Heiligenstadt, Germany) overnight. Afterwards 200 µL of the preparatory culture were
transferred into 500 mL of media (12.8 g/L disodium hydrogenphosphat-heptahydrat, 3 g/L
potassium dihydrogenphosphat, 0.5 g/L sodium chlorid, 1 g/L ammonium chloride, 0.01 g/L
calcium dichloride and 0.4 wt % glucose) and incubated for 24 hours at 37 °C. Growth of bacteria
was tracked by light absorption at 600 nm using a spectro photometer V550 Jasco Germany
GmbH, Gross-Umstadt, Germany. Bacterial growth was stopped by temperature reduction to 4
°C, centrifugation at 4500 rpm and sterile filtration (0.2 µm) of the bacterial suspension. For
conservation 0.05 (v/v) % sodium azide was added. Required endotoxin concentrations for
endotoxin recovery studies were adjusted by dilution with depyrogenated water.
2.4.2 Preparation 2
For growth of bacteria, 5 mL LB media (10 g/L sodium chloride, 5 g/L yeast extract and 10 g/L
tryptone) were inoculated with desired bacterial strain, followed by incubation at 37 °C in a
shaking incubator (Platform shaker: Innova 2300, New Brunswick Scientific Co, Enfield, USA;
Incubator: Wärmeschrank für Plattformschüttler, Mytron Bio- und Solartechnik GmbH, Heilbad
Heiligenstadt, Germany) overnight. Afterwards the preparatory culture was transferred into 20
mL of media*1 and incubated for 18 hours defined temperatures*2. Bacterial growth was stopped
by temperature reduction to 4 °C and sterile filtration (0.2 µm) of the bacterial suspension. For
conservation 0.05 (v/v) % sodium azide was added. Required endotoxin concentrations for
endotoxin recovery studies were adjusted by dilution with depyrogenated water.
25
*Modification of growth conditions:
1) LB media: a) 100 % media
b) 1 % (v/v) media in depyrogenated water
c) 100 % media plus 20 mM tris buffer
d) 100 % media plus 20 mM EDTA
2) Growth temperature: i) room temperature (20 – 25 °C)
ii) 37 °C
2.5 Sample handling
2.5.1 Addition of endotoxin to a sample
Samples were prepared in 5 mL borosilicate glass tubes with sample volumes of 1 mL per sample.
Unless otherwise described, samples were spiked with 10 µL of endotoxin from E.coli O55:B5 (gel-
filtrated) out of a 10,000 EU/mL stock solution. Before adding the endotoxin spikes to the sample,
the endotoxin stock solution was shaken at 1400 rpm for 10 minutes using Multi Reax shaker
(Heidolph Instruments GmbH & Co.KG, Schwabach, Germany). After spiking, the samples
including the endotoxin are shaken again at 1400 rpm for 60 seconds.
2.5.2 Sample preparation before endotoxin detection
To eliminate test interference and ensure validity of endotoxin detection assays, samples were
vortexed for 2 minutes at 1400 rpm and diluted in depyrogenated water immediately prior to the
measurement. The validity of the measurement was controlled by spiking of defined endotoxin
amounts to the diluted samples (Positive Product Control (PPC)). Endotoxin determination in a
sample was considered valid, if 50 to 200% of the spiked endotoxin (PPC) were recovered.
2.5.3 Incubation of endotoxin in a sample
For time dependent endotoxin recovery (hold time) studies, endotoxin was incubated in
undiluted samples over time. If not otherwise specified, the pH of used buffer systems was
adjusted to 7.5. After addition of defined endotoxins spikes to undiluted samples, the resulting
solution was vortexed for 60 seconds at 1400 rpm. Samples were subsequently stored without
further vortexing at defined temperatures (4° C (2 – 8 °C), RT (20 – 25 °C) and 37°C (35 – 40 °C)
for a desired period of time.
2.5.4 Preparation of endotoxin recovery kinetics
Samples for kinetics can be prepared in two opposed sequential arrangements: Online mode
(2.5.4.1) vs. reverse mode (2.5.4.2). Both preparations were used. Online mode has the advantage
that the endotoxin spike in all samples is exactly the same. Disadvantage is that each time point
26
needs a new test and standard curve at different days. The advantage and disadvantage
performing the reverse preparation are vice versa compared to the online mode.
2.5.4.1 Online mode – kinetics (OM)
Determination of endotoxin masking kinetics was prepared out of one stock solution using the
online mode. The start of the kinetics was defined, when at least surfactant, chelator and
endotoxin were combined and vortexed. In order to measure endotoxin at individual points of
time, 10 µL of the corresponding sample were transferred to 990 µL of depyrogenated water after
desired incubation period. Prior to the measurement, no further dilution was required. Samples
were vortexed at least for 2 minutes at 1400 rpm before the individual time points were
measured. The online method was used for all preparations of endotoxin masking kinetics, if not
otherwise indicated.
2.5.4.2 Reverse mode – kinetics (RM)
Reverse endotoxin masking kinetics was prepared by spiking aliquots of a sample at different time
points. The particular spiked and not spiked aliquoted were stored under equal conditions over
time. The aliquot with the longest endotoxin incubation period was spiked first (e.g. 7 days prior
to the measurement). Further aliquots with shorter incubation periods were spiked later in
accordance with the respective incubation period. After spiking the time point zero aliquot, all
samples were equivalently prepared and measured on the same assay.
2.5.5 Endotoxin recovery and masking controls
To control accuracy of the endotoxin spiked into the undiluted samples, equal amounts of
endotoxin were spiked into depyrogenated water (positive control), mixed and identically
incubated as the actual sample. For calculation of endotoxin recovery, the determined endotoxin
concentrations in the actual sample is compared to the determined endotoxin concentration at
time zero in the positive control and stated as percent.
2.5.6 Sample preparation for demasking of endotoxin
2.5.6.1 Preparation of demasking agents (working solution).
For demasking, different molecules (Sodium citrate, calcium dichloride, BSA, SDS, alkyl-alcohols)
and mixtures of them were used. Before addition of these molecules to the masked sample a 10-
fold concentrated working solutions of the desired component and concentration were prepared.
Irrespective of the used alkyl alcohols, the components were dissolved in depyrogenated water.
Alkyl alcohols were dissolved in 70% (v/v) ethanol.
2.5.6.2 Sample treatment for demasking of endotoxin.
Endotoxin demasking was performed by the addition of 100 μΙ of each demasking agent (2.5.6.1
working solution) to the masked sample. The particular agents were sequentially added and 2
27
minutes vortexed after each addition. After addition of all demasking agents, the samples were
incubated for 30 minutes at room temperature without vortexing.
2.6 Limulus-based endotoxin detection assays
2.6.1 Recombinant Factor C assays (rFC)
2.6.1.1 EndoZyme®
For detection of endotoxin, a recombinant Factor C test (EndoZyme ®), based on a homogenous
test format, was used according to manufacturer’s instructions. The released amount of
fluorescence substrate was measured fluorometrically at 440 nm (Excitation: 380 nm) with a
FLx800 fluorescence microplate reader (BioTek Instruments GmbH, Bad Friedrichshall, Germany).
All samples were measured in duplicate and average values were used for further calculations.
Standard curves were fit using a four parameter logistic non-linear regression model. The
detection limit of the assay was 0.005 EU/mL. EndoZyme was used in all experiments, if not
otherwise indicated.
2.6.1.2 EndoLISA®
For detection of endotoxin, a recombinant Factor C test (EndoLISA ®), based on a heterogeneous
test format, was used according to manufacturer’s instructions. The released amount of
fluorescence substrate was measured fluorometrically at 440 nm (Excitation: 380 nm) with a
FLx800 fluorescence microplate reader (BioTek Instruments GmbH, Bad Friedrichshall, Germany).
All samples were measured in duplicate and average values were used for further calculations.
Standard curves were fit using a four parameter logistic non-linear regression model. The
detection limit of the assay was 0.005 EU/mL.
2.6.2 Limulus Amebocyte Lysate assay (LAL)
For endotoxin detection using Limulus Amebocyte Lysate, a kinetic chromogenic LAL assay
(Kinetic-QCLTM) was used according to manufacturer´s instructions. The released amount of
chromogenic substrate was measured spectrophotometrically at 405 nm with an Epoch2
absorbance microplate reader (BioTek Instruments GmbH, Bad Friedrichshall, Germany). All
samples were measured in duplicate and average values were used for further calculations.
Standard curves were fit using a linear regression model. Detection limit of the assay was 0.005
EU/mL.
2.7 Silver stained SDS-PAGE Endotoxin samples (crude extracts, 2.4.1 preparation 1) were vortexed for 30 seconds. 40 µL of
the sample were mixed with 10 µL SDS sample loading buffer (1.25 µL tris buffer, 0.5 mg sodium
dodecylsulfate, 2.87 µL glycerol, 1.25 µL EDTA, 5 µg bromphenol blue and 0.25 µL
mercaptoethanol) and boiled for 10 minutes. 18 µL of each sample were loaded onto a 20%
28
gradient tris-tricin gel. The SDS-PAGE (sodium dodecyl sulfate-polyacrylamide gel
electrophoresis) was run in tris-tricine/SDS buffer at 130 V (Electrophoresis power supply: EPS
301, Amersham Pharmacia Biotec, Uppsala, Sweden; Vertical Electrophoresis unit: Mighty small
SE260, Hoefer Inc., Holliston, USA) for 90 minutes. For silver staining of the gel the following
procedure was used:
1) Fixation: Incubation of gel overnight in 150 mL of 25% (v/v) isopropanol and 7% (v/v) acetic acid.
2) Oxidation: Incubation of gel for 5 minutes in 75 mL depyrogenated water with 0.5 g of periodic acid and 1 mL of fixation solution (25% (v/v) isopropanol and 7% (v/v) acetic acid solution).
3) Washing: Four times 5 minutes wash in depyrogenated water.
4) Silver staining: 10 minutes in a solution containing 350 µL sodium hydroxide (8 M), 1 mL concentrated ammonium hydroxide (28%), 2 mL silver nitrate (20% (w/v) and 75 mL depyrogenated water.
5) Washing: Four times 5 minutes wash in depyrogenated water were done.
6) Development: 20 min in a solution containing 50 mL depyrogenated water, 50 mg citric acid and 50 µL formaldehyde (37%). Development was stopped using 10 % (v/v) acetic acid.
2.8 Calculations and plots For calculation of endotoxin recovery, plotting of graphs and simulation of endotoxin recovery
kinetics Microsoft Excel 2010, Version 14.0.7015.1000 was used. Sigmoidal experimental data
points were fit using SigmaPlot 2001 for Windows Version 7.0. Calculation of standard curves for
determination of endotoxin concentrations, Gen5 Data Analysis Software Version 2.05 from
BioTek Instruments GmbH, Bad Friedrichshall, Germany was used.
29
3 Results and Discussion
3.1 Masking of endotoxin in surfactant samples: Effects on Limulus-
based detection systems
3.1.1 Introduction
Bacteria and their breakdown products like endotoxins are ubiquitous[46]. The presence of
endotoxins in aqueous compositions is an intractable problem, which severely threatens and
limits the application of many compositions, in particular if intended for pharmaceutical use. This
is especially true for parenteral administrated biopharmaceutical drug products. Therefore, there
is a risk of endotoxin contamination in the production process of pharmaceutical drug products.
To safeguard against potentially hazardous incorporation of endotoxin, measurements must be
performed to exclude endotoxin from all steps and products used in the production process of
parenteral drug products. For such measurements, a method of choice is the LAL assay. Since
decades, these assays are positioned in quality control of pharmaceutical production and have
been proven to be a sensitive measure for endotoxins. However, some reports have shown that
detection of endotoxins is not always suitable in complex samples[47], [48]. One reason for
inadequate detection of endotoxin is interference of sample constituents with the enzymatic
reaction of the Limulus-based detection system. In this case, certain components (e.g. heavy
metals, protease inhibitors) can directly disturb enzyme activation of the detection system, which
is called test interference[49]. This phenomenon is well known and to identify test interference,
positive product controls (PPC) are performed. To this end, a known amount of endotoxin is
added to the sample and immediately measured. A test is considered valid if the spiked endotoxin
is recovered in a range of 50 to 200%. If the validity criterion is not fulfilled, it is recommended to
overcome interference by suitable sample treatments such as dilution, filtration, neutralization,
dialysis or heating[40]. Another potential reason for inadequate endotoxin detection is the
interaction of endotoxin itself with matrix components of the sample. For instance, it has been
reported that endotoxin can interact with blood components[50], proteins[47] or amphiphilic
molecules[51], [52], resulting in a significant change of endotoxin activity. Notably, approaches
which eliminate test interference problems are not effective in overcoming such effects[47]. In
the 1990´s Greaves and co-workers already differentiated between dilution-dependent and
dilution-independent interference in environmental samples[53].
In the recent past, inadequate endotoxin detection has been observed in biopharmaceutical drug
products[54]. In such cases, the APIs are mainly proteins[55], which are capable of intrinsic
binding to endotoxin as previously described by Anspach and co-workers[47]. The inadequate
detection of endotoxin might be explained by protein-endotoxin interactions, but furthermore,
30
therapeutic proteins are usually stabilized by excipients, like nonionic surfactants and certain
buffer components[56]. Surprisingly, endotoxin spiking experiments in formulations that lack the
API resulted in LER. Such observations of disturbed endotoxin determinations in
biopharmaceutical products occurred over time and the related risk of undiscovered endotoxin
contamination events compelled us to study the impact of common formulation components on
the detectability in Limulus-based detection systems.
Therefore, crucial formulation components of common biopharmaceuticals are extracted from
popular biopharmaceutical drug products (table 1) and endotoxin recovery out of such buffer
systems containing single and multiple components is investigated. Due to the expected time-
dependency of LER, the end-point of the reaction is determined by using different sample
incubation temperatures. While multi-parameter interactions between surfactants, complex
forming agents and endotoxin are assumed, the particular influence of the buffer components on
the detectability of endotoxin is focused. Hence, the impact of pH in a sample, different buffer
systems and the effects of different nonionic surfactants are studied. Finally, various endotoxin
concentrations are added to a LER causing formulation to evaluate the masking capacity of such
a sample.
31
3.1.2 Results
For analysis of LER, undiluted samples were spiked with defined amounts of endotoxin and
incubated over time. With regard to biopharmaceutical drug products, single and mixtures of
common formulation components were examined to identify critical components or
combinations affecting endotoxin detection. Endotoxin recovery in the presence of different
formulation components is shown in table 2.
Table 2 Endotoxin recovery out of single and multiple component samples
Endotoxin recovery over time in presence of single and multiple formulation components is shown. Samples were spiked with an endotoxin amount of 10000 EU/mL. Endotoxin was detected after preparation (approx. 45 min., T0)) and after sample incubation of seven days (T7) at room temperature. Prior to the measurement samples were diluted up to 1:10000.
Sample: Formulation components: T0 Recovery [%] T7 Recovery [%]
1 H2O 100 94
2 Sodium citrate 125 94
3 Sodium phosphate 95 69
4 Polysorbate 20 91 79
5 Sodium citrate + polysorbate 20 1 0
6 Sodium phosphate + polysorbate 20 52 0
The recovery was determined after endotoxin spiking of the samples without incubation (T0) and
after sample incubation for seven days (T7) at room temperature (RT). Samples 2, 3 and 4,
containing only single component additions showed no significant loss of activity over time,
according to a validity criteria of 50 to 200% of endotoxin recovery. In contrast, endotoxin could
not be detected in samples containing both buffer and surfactant (samples 5 and 6) after an
incubation period of seven days. This result shows that endotoxin recovery is substantially
affected by the sample incubation period and mixture of formulation components.
32
Figure 7 Endotoxin recovery kinetics in citrate-polysorbate formulations
The endotoxin recovery is plotted as a function of the incubation time. The different curves indicate
incubation temperatures at 35-40°C ( ● ), 20-25°C ( ) and 2-8°C ( ). 100 EU/mL endotoxin were added to samples containing 10 mM sodium citrate and 0.05 % polysorbate 20 and incubated for different time periods (reverse mode, 2.5.3.2). For detection (A) kinetic chromogenic LAL tests and (B) recombinant Factor C tests were used. The error bars reflect the standard deviation of three independent replicates (n=3) of the sample. The replicates were measured on the same microtiter plate.
0
50
100
150
200
250
0 1 10 100
Re
cove
ry [
%]
A
0
50
100
150
0 1 10 100
Re
cove
ry [
%]
Time [hours]
B
33
To examine the time-dependency of LER more in detail, kinetics of endotoxin recovery in samples
containing polysorbate 20 and sodium citrate was analyzed. Figure 7 shows the endotoxin
recovery of three identical samples as a function of time at different incubation temperatures
(4°C, RT and 37 °C) using a Limulus amebocyte lysate (LAL) test (A) and a recombinant Factor C
test (B) for detection. After approximately 24 hours of incubation, all samples showed low
endotoxin recovery in both detection systems. This result clearly indicates that this phenomenon
is independent of the test system. Furthermore, the loss of activity was significantly accelerated
with increasing incubation temperature.
Together, these experiments show that the combination of a buffer system and a surfactant
results in LER over time. Thus, the impact of different buffer systems was studied (Figure 8). In
order to investigate pH dependency of endotoxin recovery after sample incubation, defined
compositions under different pH conditions were studied (figure 8A). In the absence of
surfactants, the variation of pH had no impact on endotoxin detection. This confirmed again the
previous results (table 2). However, in the presence of polysorbate the recovery significantly
decreased at pH values higher than pH 2 (citrate system) and pH 5 (phosphate system),
respectively. The transition to higher pH values hampered endotoxin recovery. In addition, the
diverging curve progressions (Figure 8A) indicate an intrinsic effect of each particular buffer
system. Hence, endotoxin recovery kinetics using different buffer systems such as
ethylenediaminetetraacetic acid (EDTA), citrate and phosphate were studied (Figure 8B).
0
1
10
100
1 2 3 4 5 6 7 8 9
Rec
ove
ry [
%]
pH
A
34
Figure 8 Impact of buffer system on endotoxin recovery
(A) Endotoxin recovery is shown as a function of the pH. The pH varied in a range from 1 to 9 and incubation was at RT for seven days. 100 EU/mL of endotoxin were added to solutions containing 0.05
wt % polysorbate 20 plus 10 mM citrate ( ), 0.05 wt % polysorbate 20 plus 10 mM phosphate ( ),
citrate only ( ) and phosphate only ( ). The error bars reflect the standard deviation of three independent replicates (n=3) of the sample. The replicates were measured on the same microtiter plate. …………………………………………………… (B): Endotoxin recovery is plotted as function of incubation time using different buffer systems. Sample incubation was at RT. 100 EU/mL of endotoxin were added to solutions containing a buffer (5 mM EDTA
( ), 10 mM sodium citrate ( ) or 10 mM sodium phosphate ( )) and 0.05 wt % polysorbate 20 (reverse mode, 2.5.3.2). The error bars reflect the standard deviation of three independent replicates (n=3) of the sample. The replicates were measured on the same microtiter plate.
The endotoxin recovery in the described buffer systems is plotted as a function of time. The
system containing EDTA showed a rapid loss of activity. The loss of activity was slower under citric
conditions and slowest under phosphoric conditions. After six hours, recovery was below 30%
under each condition. As confirmed before, surfactants are significantly involved in reducing the
activity of endotoxin in common detection systems. Therefore, the effects of different surfactants
at constant buffer and endotoxin conditions were examined. In figure 9, the endotoxin recovery
out of surfactant solutions (polysorbate 20, polysorbate 80 and octoxynol 9) in the presence and
absence of citrate are plotted as a function of surfactant concentration. In general, all used
surfactants substantially reduced endotoxin detectability in the presence of citrate after seven
days of incubation. In absence of citrate, only octoxynol 9 showed low recovery at relatively high
concentrations after the incubation period. Without incubation, endotoxins were detectable in
all cases (> 50%), except using highest octoxynol 9 concentration including citrate.
0
1
10
100
0 1 2 3 4 5 6
Rec
ove
ry [
%]
Time [hours]
B
35
Figure 9 Surfactant dependent endotoxin recovery
Endotoxin recovery is shown as a function of the particular surfactant concentration. 100 EU/mL of endotoxin were added to solutions containing various amounts of (A) polysorbate 20, (B) polysorbate 80 or (C) octoxynol 9. Endotoxin recoveries were determined in the presence of 10 mM sodium citrate
immediately after preparation ( ) and after incubation for seven days at RT ( ). In parallel endotoxin
activities were determined in the absence of citrate, without incubation ( ) and after incubation
( ).
0
50
100
150
200
0,00001 0,00010 0,00100 0,01000
Reco
very
(%
)
Polysorbate 20 (%(w/v))
A
0
50
100
150
200
0,000001 0,000010 0,000100 0,001000
Reco
very
(%
)
Polysorbate 80 (%(w/v))
B
0
50
100
150
200
0,00001 0,00010 0,00100 0,01000 0,10000
Reco
very
(%
)
Octoxinol 9 (%(w/v))
C
36
Summarizing the results above, nonionic surfactants and complex forming buffer components in
combination cause a significant perturbation of endotoxin detection in Limulus-based detection
systems. The resulting LER is time-dependent and may solely occur after a certain period of time.
For a final evaluation, the masking capacity was examined. Endotoxin was titrated into samples
containing a citrate buffer system and polysorbate.
Figure 10 Endotoxin masking capacity of citrate-polysorbate 20 formulation
Detectable endotoxin concentration is shown in relation to the spiked endotoxin concentration. Defined amounts of endotoxin were added to solutions containing 0.05 wt % polysorbate 20 and 10 mM sodium citrate and incubated for seven days at 4 °C. Endotoxin spikes were prepared out of a LPS stock solution containing 10-6 EU/mL.
Figure 10 shows the endotoxin masking capacity of such a particular matrix. Spiked endotoxin
contents of up to 250 EU/mL resulted in no endotoxin recovery after seven days of incubation.
Medium and high-level spikes resulted in very low (<1%) and low endotoxin (< 5%) recovery. This
result illustrates a high masking capacity of common formulation matrices and suggests the need
for vigilance in BET under such conditions.
6079,5
97,425
0,97
< 0,2 < 0,2 < 0,2
0
1
10
100
1000
10000
100000
1000000
250000 25000 2500 250 25 2,5
End
oto
xin
act
ivit
y [E
U/m
L]
Endotoxin spike [EU/mL]
37
3.1.3 Discussion
BET is a standardized control instrument in pharmaceutical microbiology. To check the absence
of test interference and validity of a measurement, PPCs are regularly used according to
pharmacopoeial requirements. However, since the observation of inconsistencies during BET,
although valid PPC are given, users are partly asked by regulators to establish additional test
procedures for storing samples and demonstrating the stability of assayable endotoxin
contents[57]. Consequently, there are two different control procedures in BET. The first are PPCs,
which are used to identify test interference. Thereby, defined contents of endotoxin are spiked
into diluted samples immediately before the measurement is started. The measurement is
considered valid, when 50 to 200% of this spike is recovered. The second procedure to investigate
LER is the application of “hold time” or “endotoxin recovery” studies. In this case, defined
contents of endotoxin are spiked into undiluted samples and the spike is incubated for a certain
period of time in the sample before the actual endotoxin measurement. A sample is popularly
classified as LER, when less than 50% of the spike is recovered over time.
Basically, test interference directly affects the detection system, but can be overcome by dilution.
In the case of LER, sample dilutions up to a factor of 10,000 could not overcome inadequate
recovery (Table 2), demonstrating that LER is dilution independent and therefore not caused by
test interference. Furthermore, after short incubation periods of endotoxin in the sample, the full
endotoxin content could be recovered, showing the functionality of the detection system (Figure
7). These findings confirm a previous observation, namely that under certain conditions the
interference in Limulus-based detection methods is dilution independent. In this case, it is
assumed that the aggregate conformation of LPS is affected and not the detection system
itself[53]. Hence, the results above indicate that the phenomenon of LER is also caused by
alterations in the aggregate conformation of the endotoxin, meaning the endotoxin is masked.
This is also supported by the time-dependent appearance of LER, while test interference appears
immediately and is therefore time-independent. This time-dependent behavior of LER can be
illustrated by an alteration of the supramolecular structure of the amphiphilic LPS. In general, the
process of aggregation of amphiphilic molecules can be very variable with respect to time-scales
for structural changes, which range from sub-microseconds to several days, weeks and even
month[58]. This time-dependent occurrence of LER might also explain confounding experimental
results in pharmaceutical industries (results not published), in which the LER phenomenon was
not observed, although masking conditions were present.
3.1.3.1 Effects of complex forming agents
The results demonstrate that endotoxin recovery is affected by the formulation components
themselves, even if the active pharmaceutical substance, such as a protein, is absent. The
38
simultaneous presence of a nonionic surfactants and complex forming components (chelator)
suffices to decrease the detectability of endotoxin. The presence of only one of the formulation
components does not effectively disturb endotoxin recovery. These findings confirm former
assumptions of endotoxin disaggregation [59], [60] and explain the interdependent interaction of
surfactant and chelator on endotoxin. Due to the amphoteric and amphiphilic nature of LPS
(Figure 3), complex forming agents disturb the electrostatic interactions and surfactants
potentially disturb the hydrophobic interactions in endotoxin aggregates. Certainly, to disturb the
supramolecular structure of endotoxin a reduced rigidity is beneficial. The rigidity is controlled by
the salt form of LPS, which again involves the presence of multivalent cations like Ca2+.[61], [62]
Consequently, it can be assumed that complex forming agents are in competition with negatively
charged patches of the endotoxin. Therefore, the salt bridges between LPS molecules are
disturbed, which should result in a reduced rigidity of endotoxin aggregates, which in turn
facilitates changes in the supramolecular structure. In the presence of EDTA the recovery drops
faster than in the presence of citrate or phosphate based buffer systems (Figure 8B). Using buffer
components with higher metal complex forming capabilities accelerate masking kinetics. Thus,
the chelating capability of the buffer component is crucial. The related metal complex formation
constants are directly proportional to the denticity of the ligand (rule of thumb[63]). A
hexadentate ligand like EDTA forms more stable metal complexes than a tridentate ligand like
citrate. Furthermore, the equilibrium complex formation ability and the complex stability of a
chelator are pH dependent. At low pH values, protons are in competition with cations, which
hamper formation of metal complexes[64]. Consequently, LER is affected by the free
concentration of protons reflected by pH, which is controlled by the buffer system and its
particular acid dissociation constant. This explains the pH dependent endotoxin recovery in
different buffer systems (Figure 8A).
3.1.3.2 Effects of surfactants on LER
As shown above, complex forming components alone do not result in inadequate endotoxin
detection, additional amphiphilic components like surfactants are necessary. Due to the fact that
LPS itself are amphiphilic, these tend to aggregate, driven by the low solubility of hydrophobic
fatty acids of lipid A in an aqueous solution[65]. Thus, LPS exhibit certain supramolecular
structures, which influence detectability in Limulus-based detection systems[66]. Structural
transitions of amphiphilic systems are affected by a large variety of physical and chemical
parameters. One major principle to control these structures is the head group repulsions of self-
assembling molecules. They can be affected by co-surfactants, electrolytes, and amphiphilic
counter ions[67]. If the masking surfactant (e.g. polysorbate) intercalates between LPS molecules
resulting in reduction of head group repulsions, the establishment of a new equilibrium is favored
39
and the supramolecular structure of LPS is altered. The interaction of nonionic surfactants with
LPS aggregates is favored if the LPS aggregates possess a certain degree of rigidity. The latter is
controlled, to some extent, by ionic interactions as described above. To this end, after the
addition of surfactants and chelators to LPS solutions, the supramolecular structure of LPS is
changed into a structure with a lower affinity to the endotoxin sensitive Factor C of the Limulus-
based detection system resulting in the measurement of a lower activity. Such a structure could
be disaggregated LPS due to a molecular excess of surfactants. This hypothesis fits well to the
observation of Mueller et al., who have shown that disaggregated LPS molecules (“monomers”)
are substantially less active than aggregated LPS in the detection system used[66]. Additionally,
Tan et al. proposed a cooperative binding mechanism of LPS to Factor C, which consequently
requires more than one LPS molecule in close spatial arrangements[68]. On the other hand, it
has been shown that monomeric LPS show a higher potency in activating LAL assays than
aggregated LPS[27]. Under these circumstances, the inadequate detectability might has steric
reasons, in which the activating spots of the LPS (lipid A) are hidden by surfactant molecules and
are not accessible for Factor C.
3.1.3.3 Potential two-step reaction mechanism of LER
In summary, we propose a two-step mechanism of endotoxin masking. Figure 11 illustrates the
effects of chelating buffer components and nonionic surfactants on endotoxin. In this mechanism
the equilibrium LPS structure is shifted to an altered supramolecular structure.
Figure 11 Two-step mechanism of endotoxin masking
Potential reaction mechanism of endotoxin masking, caused by complex forming agents and surfactants is schematically illustrated. In a first step, pure endotoxin aggregates are disturbed by chelators reducing the rigidity of the aggregate. Then, surfactants interact with endotoxin by forming mixed aggregates.
In its natural state, LPS monomers tend to aggregate due to the hydrophobic interactions
between the lipid A molecules. Additional ionic interactions formed by divalent cations and
40
negatively charged substitutes (e.g. phosphates) of the LPS increase the rigidity of aggregates. By
adding a complex forming agent (I), the salt bridges formed by divalent cations (e.g. Mg2+) and
LPS are destabilized, leading to a reduced rigidity of the aggregate. The additional presence of a
surfactant (II) can then change the initial supramolecular structure by formation of mixed
aggregates. This structural change leads inevitably to a certain change in detectable activity, as
endotoxin activity is dependent on its supramolecular structure. Due to the common molar excess
of complex forming agent and surfactant (micromolar range) compared to endotoxin content
(nanomolar range), mixed surfactant micelles containing monomerized LPS are the most probable
resulting supramolecular structure.
Together, this study confirmed the phenomenon of LER in Limulus-based detection systems and
exemplifies a potential mechanism of endotoxin masking. Due to the time-dependency of
masking, the unknown period of endotoxin presence during a potential event of endotoxin
contamination in a sample defines the chance of endotoxin recovery. Hence, LER is under control
of kinetics that governs the formation of mixed endotoxin-surfactant aggregates, which make
them less prone to activate the enzymatic reaction of Limulus-based detection systems. Capacity
experiments have shown that commonly used excipients are capable of masking hazardous
amounts of endotoxin. Consequently, the phenomenon of LER has to be especially considered
during quality control of drug products including surfactant and complex forming agents. In order
to further investigate the time-dependent masking behavior of endotoxin, detailed reaction
kinetics is studied in the next chapter (3.2).
41
3.2 Endotoxin Masking: A kinetically controlled reaction mechanism
3.2.1 Introduction
In the previous chapter (3.1), the phenomenon of LER was studied in common biopharmaceutical
product matrices and it is supposed that LER is caused by the interaction of sample matrix and
endotoxin, resulting in masking of endotoxin. Basically, LPS aggregate due to their amphiphilic
nature and in turn form certain supramolecular structures[70]. Yet, during the transition from
detectable to undetectable (masked) endotoxin a change in its supramolecular structure is
probable. Moreover, a disaggregation of LPS may occur during masking. With regard to parenteral
drug products, regulators consider this phenomenon as a potential safety issue due to the
possible underestimation of critical endotoxin levels in a sample[54]. To investigate if a product
is affected by LER, an industrial guideline for BET suggests hold time studies[57], which intend to
incubate known amounts of endotoxin over time in undiluted products prior to the actual test
procedure.
The aim of such hold time studies is to prove assayability of endotoxin in a particular sample over
time. In order to provide a better understanding of the LER mechanism and in turn to improve
efficiency in planning of such hold time studies the time-dependent appearance of LER is analyzed
in detail. Therefore, a common formulation matrix containing sodium citrate and polysorbate 20
is used in the present study. This matrix was chosen, because it is a common formulation
composition for biopharmaceutical drugs products[56]. Furthermore, such a composition reflects
the minimum factors responsible for endotoxin masking and should therefore help to elucidate
the driving forces of endotoxin masking. Apart from the temperature dependency of a reaction,
the change in concentrations during chemical reactions is often directly proportional to the rate
of a reaction[71]. Derivation of a rate law according to the underlying masking reaction enables
prediction of the reaction rate depending on a given product formulation and help to plan sample
hold time periods for identification of LER. In order to determine a rate law of the reaction,
endotoxin recovery kinetics is recorded using different concentrations of citrate, polysorbate and
endotoxin. The variation of concentrations is used to identify whether there are certain reactants
controlling the reaction rate. Furthermore, the rate determining step in the proposed two-step
reaction mechanism (3.1.3.3) is specified and the derived reaction law is and used for the
simulation of endotoxin masking kinetics.
42
3.2.2 Results
Endotoxin masking, caused by the simultaneous presence of surfactants and complex forming
agents, has been shown in the previous chapter (3.1). In these studies, mixtures of formulation
components influenced the occurrence of masking. To analyze whether the preparation of the
samples affects endotoxin masking, kinetics with different order of sample preparations was
investigated. Two of the three components (polysorbate, sodium citrate and endotoxin) were
pre-incubated overnight and masking kinetics was started by addition of the third component
(Figure 12). As expected, all preparations were affected by LER, but diverging kinetics are
observed. Endotoxin pre-incubated with sodium citrate shows the fastest masking kinetics and
pre-incubation of polysorbate with sodium citrate shows the slowest kinetics. Endotoxin pre-
incubated with polysorbate shows likewise slow kinetics. Based on the accelerated reaction
kinetics by pre-incubation of endotoxin with sodium citrate, the interaction between endotoxin
and sodium citrate appears to be the time limiting reaction step. The large error bars reflect test
variables of Limulus-based assays, but also the experimental setup. Depending on exact ambient
temperatures and hands on time for sample preparation including spiking, mixing as well as
vortexing may influence reaction kinetics.
Figure 12 Endotoxin recovery depending on order of matrix component and LPS addition
Endotoxin recovery is plotted as a function of incubation time. 100 EU/mL endotoxin were spiked to solutions containing 0.05 wt % polysorbate 20 and 10 mM sodium citrate. The particular kinetics was generated by different sequential arrangements during sample preparation at RT. In the first kinetics
(dark grey triangles ( )) LPS and sodium citrate were pre-incubated (overnight) and polysorbate was added at time zero (0 min) to start the reaction. In the second kinetics (light grey squares ( )) LPS and polysorbate were pre-incubated (overnight) and sodium citrate was added at time zero (0 min). In the
third kinetics (black diamonds ( )) sodium citrate and polysorbate were pre-incubated (overnight) and LPS was added at time zero (0 min). For calculation of the data points the mean values of two (LPS/sodium citrate and LPS/polysorbate pre-incubation) and three (polysorbate/sodium citrate pre-incubation) individually performed repetitions of the kinetics were used and the error bars reflect the corresponding standard deviations. For a better comparison of independent measurements, the data was normalized and the starting points were set to 100 %.
0
20
40
60
80
100
120
0 10 20 30 40 50 60
Reco
very
[%
]
time [min]
43
In order to further analyze the driving forces of this endotoxin masking effect, recovery kinetics
using different concentrations of the reactants were analyzed. First of all, recovery of different
endotoxin concentrations (50, 500 and 5000 EU/mL) were studied under constant
polysorbate/citrate conditions (Figure 13). The recovery over time showed no significant
difference using different endotoxin concentrations, indicating that masking is independent of
the initial endotoxin concentration. After 10 minutes of incubation, all recoveries are above 50%
and after 45 minutes all recoveries are below 7%.
Figure 13 Endotoxin recovery kinetics depending on LPS concentration
Endotoxin recovery is plotted as a function of incubation time. Varying contents of endotoxin were added to samples containing 0.05 wt % polysorbate and 10 mM sodium citrate at RT. The black columns reflect 5000 EU/mL, grey columns 500 EU/ml and white columns 50 EU/mL. For calculation of the data points the mean values of two individual prepared kinetics were used and the error bars reflects the corresponding standard deviations. For a better comparison of independent measurements, the data was normalized and the starting points were set to 100 %.
Thereafter, masking of endotoxin was analyzed using reduced concentrations of polysorbate and
sodium citrate (Figure 14A). The recovery of endotoxin in a sample containing 0.05 wt %
polysorbate and 10 mM sodium citrate was below 1 % after one hour of incubation. Using a
sample matrix of 0.0125 wt % polysorbate and 2.50 mM sodium citrate the endotoxin recovery
was not reduced after one hour of incubation, but after incubation of 20 hours, the endotoxin
recovery was very low, too. Using a sample matrix containing 0.0008 wt % polysorbate and 0.16
mM sodium citrate, no significant decrease in recovery of endotoxin is observed over the analyze
period of time (20 hours). Thus, masking can be delayed or even avoided, when the entire sample
matrix is diluted before spiking the endotoxin. Furthermore, endotoxin recovery was determined
in samples, in which only the polysorbate concentration (0.0500, 0.0125 and 0.0008 wt %) was
reduced (Figure 14B).
0
10
20
30
40
50
60
70
80
90
100
0 5 10 20 30 45 60 120 180
Re
cove
ry [
%]
Time [min]
44
Figure 14 Endotoxin recovery kinetics depending on concentration of matrix components
A) Endotoxin recovery is plotted as a function of incubation time. 100 EU/mL of endotoxin were spiked to samples containing polysorbate 20 and sodium citrate at RT. The three different colored columns reflect different polysorbate and sodium citrate concentrations. The set of black columns correspond to 0.0500 wt % polysorbate and 10.00 mM sodium citrate, the set of grey columns correspond to 0.0125 wt % polysorbate and 2.50 mM sodium citrate and the set of white columns correspond to 0.0008 wt % polysorbate and 0.16 mM sodium citrate. …………………………………………… B) Endotoxin recovery over time in samples containing polysorbate, sodium citrate and LPS is shown. The different colored columns reflect different polysorbate concentrations. The set of black columns correspond to 0.0500 wt % polysorbate, the set of dark grey columns correspond to 0.0125 wt % polysorbate and the set of light grey columns correspond to 0.0008 wt % polysorbate. The concentrations of endotoxin (100 EU/mL) and sodium citrate (10 mM) as well as temperature (RT) were kept constant.
Concentrations of the other components (endotoxin and citrate) were kept constant. In this case,
the recovery kinetics is similar within a given measurement uncertainty. The endotoxin recovery
is below 2 % independent of the polysorbate concentration after one hour of incubation. This
result differs from the previous result, in which the endotoxin was recovered after one hour of
incubation (> 100%) at reduced polysorbate and citrate concentrations (Figure 14A). Although the
polysorbate concentration was comparably titrated in both cases, the sodium citrate
concentrations were different among the two experiments (Figure 14: A vs. B). This again
indicates that the role of sodium citrate is crucial within the given experimental conditions. Due
to the fact that sodium citrate is capable of forming metal complexes with divalent cations,
endotoxin recovery kinetics was examined in the presence of divalent cations (Figure 15A). Here,
masking kinetics is delayed in the presence of 1 mM magnesium dichloride compared to the
samples without magnesium dichloride. In the presence of 5 mM magnesium dichloride, no
0
20
40
60
80
100
120
140
160
0,0 0,1 0,2 0,3 0,5 1,0 1,5 2,0 3,0 20,0
Reco
very
[%
]
time [h]
A
0
20
40
60
80
100
120
140
160
0 10 20 30 45 60
Reco
very
[%
]
time [min]
B
45
reduced endotoxin recovery is observed within the analyzed time scale. Importantly, the
magnesium dichloride was added to the samples prior to endotoxin. In a further experiment,
magnesium dichloride was added to the polysorbate/sodium citrate matrix 20 minutes after start
of the reaction (figure 15B). Although the recovery of endotoxin was already reduced at this point
of time, no further decrease of endotoxin recovery was observed after addition of magnesium
dichloride. In comparison, the sample without supplementary addition of divalent cations was
masked as expected. Hence, the addition of magnesium dichloride can stop endotoxin masking
and keep the recovery constant at the actual level. Notably, the original endotoxin activity could
not be retrieved after the addition of magnesium dichloride.
Figure 15 Endotoxin recovery depends on the presence of divalent cations
A) Endotoxin recovery is plotted as a function of time in samples containing polysorbate 20, sodium citrate, endotoxin and magnesium dichloride. The different colored columns reflect different contents of magnesium dichloride (0 mM (white columns), 1 mM (grey columns), 5 mM (black columns)). Concentrations of polysorbate (0.05 wt %), sodium citrate (10 mM) and endotoxin (100 EU/mL) as well as temperature (RT) were kept constant. ……………………… B) Endotoxin recovery is plotted as a function of time in a sample containing 0.05 wt % polysorbate, 10 mM sodium citrate and 100 EU/mL endotoxin (white columns). After 20 minutes, the sample was divided into two equivalent aliquots (1 mL each), whereof one aliquot was treated once by the addition of 20 mM (20µL of 1 M) magnesium dichloride and the other fraction were continued without treatment. Endotoxin recovery of the treated fraction is expressed by the black columns.
0
20
40
60
80
100
120
140
160
0 30 60 120
Reco
very
[%
]
time [min]
A
0
20
40
60
80
100
120
140
160
0 5 10 15 20 30 45 50 60 80 90 110 120
Recv
oery
[%
]
time [min]
B
46
Obviously, there is a relation between masking kinetics and complex formation. For deeper
analysis of this effect, masking kinetics using three different citrate concentrations (5, 20, and 80
mM) were recorded (Figure 16). A distinct acceleration of masking by increasing the citrate
concentration in the sample from 5 mM to 80 mM is observed. Furthermore, the experimental
data points are compared to a model curve using an exponential decay function, in which the
endotoxin recovery [LPSd] is calculated as a function of time t:
[LPSd]=[LPSd]0*exp(-[Ci]kt) (1)
The parameters of this function were set in analogy to the experimental conditions. Endotoxin
recovery at time 0 [LPSd]0 was set to 100%, time t was given in minutes and k was chosen by
chance and set to 7. The sodium citrate concentrations [Ci] were set to 0.005, 0.020 and 0.160
mol/L in the particular curves. The simulated curves are in good agreement with the experimental
data, illustrating that the endotoxin masking kinetics is depending on the sodium citrate
concentration. Derivation of this model function is discussed in chapter 3.2.3.1 Simulation of LER
kinetics.
Figure 16 Sodium citrate concentration determines endotoxin recovery
Endotoxin recovery is plotted as a function of incubation time. 100 EU/mL of endotoxin were added to samples containing varied sodium citrate concentrations (5 mM ( ), 20 mM ( ), 80 mM ( ). Concentration of polysorbate 20 (0.05 wt %) and temperature (RT) were kept constant. For the corresponding simulation ( ), an exponential decay function was used (Equation 1). For a better comparison of independent measurements, the experimental data was normalized and the starting points were set to 100 %.
0,0
20,0
40,0
60,0
80,0
100,0
120,0
0 20 40 60 80 100
Re
cove
ry [
%]
time [min]
47
Summarizing the kinetics above, 100 EU/mL endotoxin will be substantially masked within 90
minutes of sample incubation in the presence of at least 5 mM sodium citrate and 0.05 wt %
polysorbate 20. Variation of the polysorbate concentration (Figure 14B) showed no significant
acceleration or deceleration of the masking kinetics. In contrast, the increase of sodium citrate
concentration (Figure 16) resulted in considerably accelerated endotoxin masking kinetics. To
determine, whether there is a minimum citrate concentration for endotoxin masking, sodium
citrate was titrated under constant polysorbate and endotoxin concentrations and the samples
were incubated for seven days prior to endotoxin measurement (Figure 17). According to the
previous kinetics, it was assumed that equilibrium of the masking reaction will be established
after seven days. Plotting endotoxin recovery as a function of citrate concentration results in a S-
shaped data point’s progression, from full recovery at a constant level to no recovery of endotoxin
depending on the citrate concentration. This behavior indicates that there is a limiting
concentration of citrate to facilitate endotoxin masking which can be deduced from the transition
point. In order to determine this citrate concentration, the experimental data set was fitted using
the following nonlinear fit function:
y(x)=a/(1+(x/x0)b) (2)
Figure 17 Endotoxin recovery depends on sodium citrate concentration under equilibrium conditions
Endotoxin recovery is plotted as a function of sodium citrate concentration. 50 EU/mL endotoxin and 0.05 wt % polysorbate were incubated with varying sodium citrate concentrations for seven days at RT prior to the measurement. For calculation of the data points the mean values of two individual performed repetitions were used. Error bars reflect standard deviations. For the corresponding fit a logistic function with three parameters is used (Equation 2).
2D Graph 7
Sodium Citrate [mM]
0,001 0,01 0,1 1 10
0
50
100
150
200
Re
cove
ry [
%]
48
A three parameter logistic function (Equation 2) was chosen, because it reflects a sigmoid curve
progression and the transition point can be determined easily. The resulting sigmoid curve fit
shows endotoxin recovery as a function of sodium citrate concentration. The calculated
coefficients “a” equates 125.66, “b” equates 1.50 and “x0“ equates 0.06. The transition point (x0)
of the curve can be assumed as the limiting citrate concentration and corresponds to a citrate
concentration of 0.06 mM under given conditions (Figure 17). In consequence, masking of
endotoxin does not occur in solutions containing polysorbates and less than 0.06 mM sodium
citrate.
49
3.2.3 Discussion
Endotoxin masking (LER) has been identified as a time-dependent phenomenon and in contrast,
test interference occurs instantly[72], [73]. The latter can be therefore clearly distinguished from
masking. The results presented here show that endotoxin masking in a sample may not be
discovered when the spike is added to the diluted sample, because the original ability of the
sample matrix for masking may be weakened or even avoided, when the concentration of a matrix
component is reduced. For example, Figure 14A shows a significant delay in time until LER is
recognized when endotoxin is spiked in a diluted sample. Due to the fact that test interference is
considered since decades, common and compendial BET procedures are trimmed to identify and
overcome test interference, but these procedures are not prone to discover and overcome time
dependent masking effects. Therefore, it is important to universally include endotoxin spikes into
undiluted samples to actual test procedures as well as careful consideration of suitable incubation
conditions. In 2012, US FDA has already updated their guidelines for BET[57] and European
Pharmacopoeia includes hold time experiments in the coming 9th edition (2017). To this end, it is
necessary to generally extend and harmonize worldwide compendial test procedures to
thoroughly identify the endotoxin masking capability of a sample.
However, to control the phenomenon of LER, understanding of the masking mechanism is a key
factor. There are several examples showing that endotoxins interact with a variety of
components, including proteins[47], surfactants[51] or nano particles[74], but the reaction
mechanism remains to be elucidated. Most likely, due to the amphiphilic and amphoteric
character of LPS[9], hydrophobic and electrostatic interactions are involved. According to
previous assumptions the two-step masking mechanism (3.1.3.3) may formally be described as
follows:
[M-LPS]+[Ci] [LPS]+[M-Ci] (3)
Chelating buffer components (Ci) destabilize salt bridges formed between divalent cations (M)
(e.g. magnesium) and negatively charged substitutes (e.g. phosphates) of LPS (LPS) (Equation 3).
[LPS] +[P] [P-LPS] (4)
Subsequently, non-ionic surfactants (P) (e. g. polysorbate) can interact with LPS and result in an
altered supramolecular structure of LPS (P-LPS) (Equation 4), leading to a change in detectable
activity. For a deeper understanding of this mechanism, identification of the time depending
reaction step is necessary. To this end, the kinetics of endotoxin masking was studied in the
presence of citrate and polysorbate. Interestingly, pre-incubation of LPS with citrate and the
subsequent addition of polysorbate resulted in very fast masking kinetics. However, pre-
50
incubation of LPS with polysorbate, and subsequent addition of citrate resulted in somewhat
slower kinetics (Figure 12). Furthermore, masking kinetics with increased citrate concentrations
and constant polysorbate concentrations (Figure 16) show accelerated reaction rates. In turn, a
variation of polysorbate concentration under otherwise identical conditions had no significant
impact on the reaction rate (Figure 14B). As a consequence, the first step of the reaction
mechanism (Equation 3) seems to control the reaction rate and depends on the citrate
concentration.
3.2.3.1 Simulation of LER kinetics
To establish a simplified model describing the reaction rate, only the first step of the reaction
(Equation 3) will be considered. The second step of the reaction (Equation 4) can be neglected,
because this step is fast and does not limit the reaction rate within the given conditions. Basically,
the reaction rate [R] is given by the change of detectable endotoxin [LPS d] as a function of time t
and can be expressed as follows:
R = d[LPS d] /dt (5)
In addition, the results implicate that the reaction rate depends on the citrate concentration
(Figure 16). Generally, it is supposable that the reaction rate is also depending on the LPS
concentration, although the results indicate no concentration dependency of LPS (Figure 13).
Thus, the reaction rate R of endotoxin masking is described as depending on detectable LPS [LPS
d] and the citrate concentration [Ci] resulting in the following equation:
R = k[LPS]d[Ci] (6)
Equalizing equation (5) and (6) results in a differential function, which is the basis for a second
order reaction kinetics:
d[LPS d] /dt = k[LPS d] [Ci] (7)
After rearrangement and integration of equation (7), the rate equation of a second order reaction
is obtained, provided that the concentrations of LPS [LPS d] and citrate [Ci] are not equal[75]:
(1/([Ci]o−[LPSd]o)*(ln([Ci]/[Ci]o)/ ([LPSd]/[LPSd]o))=kt (8)
Yet, the change of detectable [LPSd] is based on the change of activity, which is usually given in
EU/mL, whereas citrate is given in mol/L. In order to convert EU/mL in mol/L it is assumed that 1
EU correlates approximately to 10-10 g (100 pico gram) LPS from E.coli[76], [77]. With a molar
mass of approximately 10000 g/mol for LPS, 100 EU/mL are equivalent approximately to 10-9
mol/L[78], [79]. As a consequence, 10-3 mol/L citrate is in a substantial molar excess compared to
51
10-9 mol/L of LPS ([C]>>[LPS]). Theoretically, this would already lead to reaction kinetics of pseudo
first order, because the concentration of citrate will not change significantly during the reaction
and can therefore be neglected. However, our results (Figure 16) show that the actual citrate
concentration has indeed a significant effect on the reaction kinetics. Furthermore, the results
indicate that the kinetics seem to be independent of the LPS concentration, which is reasonable,
because citrate is in an excess of up to seven orders of magnitude [C]>>[LPS]. A change of the LPS
concentration, for example by a factor of 1000, will presumably not affect the kinetics, since
citrate would still be in excess. Keeping this in mind, the starting molar concentration of citrate
[Ci]o will only be marginally reduced by subtraction of the initial molar concentration of LPS [LPSd]o
and leading to the following approximation:
[Ci]o−[LPS]o = [C]o (9)
Concomitantly, the marginal consumption of [Ci] due to the low molar concentration of LPS is
also negligible and allows for the following assumption:
[Ci]/[Ci]o = 1 (10)
With respect to the equations (9) and (10) the equation (8) can be approximated and written as
follows:
(1/([Ci]o)*(ln(1/([LPSd]/[LPSd]o))=kt (11)
Finally, the equation can be rearranged to give the detectable concentration of LPS [LPSd] as
function of time, and depending on the citrate concentration:
[LPSd] = [LPSd]o*exp(-[Ci]kt) (12)
The simulation of masking kinetics with different citrate concentrations, using equation 12,
resulted in a good agreement with the experimental data (Figure 16). Consequently, this model
reflects very well the observed behavior of the reaction, in which the endotoxin recovery is
strongly dependent on the citrate concentration. However, specifications may change, if initial
concentrations of the components are substantially changed or if additional components (e.g.
salts and proteins) are included in the sample conditions.
3.2.3.2 Minimum citrate concentration
Citrate has been identified to control the reaction kinetics of endotoxin masking. To get a deeper
understanding of the role of citrate, the minimal concentration of citrate required to initiate
masking at constant polysorbate concentrations was analyzed. The result shows that
approximately 0.06 mM of citrate is necessary for masking (Figure 17). This concentration is
52
orders of magnitude higher than the molar concentration of LPS, assuming that there is no
reasonable reaction stoichiometry. Nevertheless, citrate might destabilize LPS aggregates by
considering the potential role of citrate during masking. For example, permeabilization (reduction
of rigidity) of LPS aggregates occurs when ionic interactions between LPS molecules are disturbed.
Primarily, magnesium as well as calcium cations (M2+) stabilize LPS-LPS interactions by the
formation of salt bridges[62], [80]–[82]. It has also been shown that complex forming agents can
permeabilize such structures [83], [84]. In the presence of citrate, complex formation between
divalent cations and chelator occurs. Thus, it is supposed that citrate competes for divalent
cations bridging LPS molecules resulting in a permeabilization of LPS aggregates, because of
complex formation[85]. Corresponding complex formation constants of magnesium or calcium
citrate are in the range around 0.05 mM[86]–[88]. This might explain the required molar excess
of citrate compared to LPS for masking, because at lower citrate concentration, the complex
formation of calcium or magnesium by citrate is not favored and in consequence the LPS-LPS salt
bridges are not destabilized, which prevents intercalation of surfactants and in turn masking of
endotoxin.
3.2.3.3 The role of divalent cations
As described above, divalent ions play an important role in the stabilization of supramolecular
LPS structures. Aggregates of LPS possess a certain degree of rigidity, maintained by salt bridges
between LPS molecules, which in turn affect the susceptibility to masking. To this end, masking
rate can be inhibited by the supplementary addition of divalent magnesium ions. This explains
former observation showing that under certain circumstances, endotoxin activity in LAL can be
maintained through the suppression of aberrant aggregation of endotoxin by saline and
buffers[89]. It is most likely that the addition of divalent cations neutralizes the complex
formation capability of chelators and favors the stabilized LPS state, because the state of
equilibrium is changed, when the sample matrix is changed. Importantly, the retrospective
addition of divalent ions to samples in which the endotoxin is masked, enabled no recovery of
endotoxin, but the progress of masking can immediately be stopped (Figure 15B). Thus, the
effective addition of cations has to take place before endotoxin is already masked. These results
demonstrate that the destabilization of LPS (Equation 3) can be prevented by the addition of
divalent ions, but when LPS is already masked, the consecutive reaction step (Equation 4) is not
affected by supplementary addition of divalent ions. To achieve a reversal of endotoxin masking,
the merely addition of divalent ions is not sufficient. For such an approach, also surfactants have
to be neutralized or removed. This approach will be discussed in detail in chapter 3.4 demasking
endotoxins.
53
3.2.3.4 Control of reaction rate
In general, to start the reaction of endotoxin masking, a certain energy barrier has to be
overcome, which controls the reaction rate. Obviously, the reaction rate can be manipulated by
its energy input. For instance, the reaction kinetics can be accelerated chemically by e.g.
increasing chelator concentration (3.2) or physically by e.g. increasing incubation temperature of
a sample (3.1). Moreover, it is conceivable that the degree of the energy barrier depends on the
endotoxin itself. Endotoxin from different sources may have different molecular structures and
in turn different masking susceptibilities due to varying stabilization mechanisms of bacterias´
outer membrane. In consequence, endotoxins with different stabilizing mechanisms may have
different energy barriers and result in manipulated reaction kinetics. To this end, the
heterogeneity of endotoxins and their detectability in LER-affected samples is studied in the next
chapter (3.3).
54
3.3 Heterogeneity of endotoxins and their detectability in common
biopharmaceutical formulations
3.3.1 Introduction
LPS play an important role in the pathogenesis and manifestation of Gram-negative infections
and in particular of septic shock. Due to the fact that the structures of LPS can vary significantly
in its O-antigen, core region and lipid A, not all endotoxins possess the same toxicity[17].
However, to control a potential contamination event in drug products by bacterial endotoxins,
the sensitive and specific detection of LPS is of great importance. BET has been proven to
effectively detect LPS. These tests are based on reactions between the lipid A of LPS and specific
enzymes derived from the clotting cascade in horseshoe crabs[33], [90]. For quantitative
detection of endotoxin, measures of the unknown samples are compared with standard curves.
Such standard curves are prepared by known and defined standard endotoxins. The primary
standard in BET is called Reference Standard Endotoxin (RSE), which is endotoxin from E.coli
O113:H10:K negative. The RSE is purified in several steps including hot phenol-extraction,
alcoholic precipitation, enzymatic digestion of nucleic acids and intense dialysis[91]. For a better
handling lactose and polyethylene glycol are added. This standard is worldwide accepted and sets
the baseline for secondary or tertiary standards. Due to the limited availability of the RSE, vendors
of BET systems distribute secondary standards called Control Standard Endotoxins (CSE), which
are calibrated against the RSE. The preparations of these secondary standards are inspired by
RSE, but the source can differ from E.coli O113 and exact production processes and formulations
are not published.
After public recognition of the LER phenomenon, many not publicly accessible endotoxin recovery
studies in biopharmaceutical drug products and their formulations were performed by endotoxin
test providers, contract labs and pharmaceutical companies. The masking effects have been
widely confirmed, especially if the tested drug products contain surfactants and chelators. Such
hold time experiments are usually performed using standardized endotoxins like CSE and RSE as
spike. However, since bacterial endotoxin tests are commercially used, the source and
preparations of appropriate standard endotoxins are debated[24]. Especially in the case of LER
the source of endotoxin can be discussed, again. It is questionable, whether LER is depending on
the endotoxins used in BET. Obviously, depending on the source, preparation and degree of
purification, the LPS itself and the accompanying matrix components can vary in a respective
preparation[11], [92]. For instance, endotoxins from different bacteria can differ in their
molecular structures. There are differences in the lipid A (e.g. acylation), core region (e.g.
substation of sugar units) and O-antigen (e.g. distribution of sugar units). Moreover, depending
55
on the preparation endotoxin suspensions may vary in their compositions. For example, crude
suspension of bacteria, popularly known as Naturally Occurring Endotoxin (NOE), may contain
substantial amounts of lipids and proteins and phenol-extracted endotoxin preparations may only
contain limited contents of hydrophobic matrix components. Some experiments have supposed
that the detectability of selected endotoxins in complex samples might be more robust compared
to detectability of commercially available standard endotoxins[93], [94]. Therefore, endotoxins
from different bacteria, grown under manipulated conditions, crude and highly purified
endotoxins as well as endogenous endotoxins are analyzed with regard to their detectability in a
polysorbate/citrate matrix in the present study.
56
3.3.2 Results
In order to study the masking susceptibility of endotoxins from different bacteria, crude
suspensions of bacterial endotoxin were prepared. To this end, bacteria from E.coli, E.cloacae,
S.marcescens, P.aeruginosa, B.cepacia, S.maltophilia and R.pickettii were grown under equal and
defined conditions (2.4.1 preparation 1). After approximately 18 hours of growth, absorbance of
all bacterial suspensions were determined (Table 3). Bacteria from E.coli and E.cloacae showed
highest absorbance (> 1.7), whereas S.maltophilia and B.cepacia showed lowest absorbance (<
0.7). Bacteria from S.marcescense showed a medium absorbance of 1.4.
Table 3 Growth of different bacteria and release of endotoxin
Growth of different bacteria and activity of their endotoxins are shown. Bacteria were grown under constant conditions and before harvesting, absorbance (600 nm) of the entire bacterial suspension were determined (2.4.1 preparation 1). Endotoxin activity was determined in the particular sterile-filtrated supernatants of bacterial suspensions.
Source Absorbance
[600 nm] Activity [EU/mL]
E.coli O55:B5 1.9 146,174
E.coli O113 1.7 402,789
E.cloacae 1.7 189,103
S.marcescens 1.4 116,175
P.aeruginosa 0.9 8,595
B.cepacia 0.7 357
S.maltophilia 0.3 4,557
R.pickettii 0.8 77,815
The absorbance reflects the evolved biomass and in turn is an indicator for growth of the
particular bacteria under given conditions. This result clearly indicates varying growth
characteristics of different bacteria under given conditions. Endotoxins are usually incorporated
in the bacterial cell wall, but during growth, substantial amounts of endotoxin are released into
the environment of the bacteria. Therefore, the endotoxin activity in the supernatant is of great
interest. Interestingly, the detectable endotoxin activities of the different bacterial supernatants
ranged from approximately 400 to 400,000 EU/mL. Between the generated biomass of cells and
the detectable activity only a weak correlation is given.
57
Figure 18 SDS-PAGE of crude bacterial endotoxin preparations from different bacteria
Silver-stained SDS-PAGE of sterile filtrated supernatants of bacterial suspensions (2.4.1 preparation 1) is shown. The lanes reflect endotoxins from E.coli O55:B5 (1), E.coli O113 (2), E.cloacae (3), S.marcescens (4), P. aeruginosa (5), B.cepacia (6), S. maltophilia (7) and R.pickettii (8).
Furthermore, these bacterial supernatants were applied to SDS-PAGE and silver stained (Figure
18). The typical ladder pattern of LPS can be observed at most lanes and the limited intensities of
the bands reflect low concentrations, which are approximately in agreement with the detected
activities (Table 3). Upon closer examination, also variations in the arrangement of bands
between the different endotoxin samples can be observed, which reflect different molecular
structures and proof heterogeneity of endotoxins. To study the masking susceptibilities of crude
endotoxin preparations, these preparations were used as endotoxin source for recovery
experiments in polysorbate/citrate matrices (Table 4). The crude extracts of endotoxins from
E.coli, E.cloacae and S.maltophilia resulted in low recovery already at day 0. Endotoxins from
S.marcescens, P.aeruginosa, and R.pickettii showed a gradual loss of recovery over time. The
endotoxins from B.cepacia could be detected over time and showed no trend in reduced activity.
This result clearly indicates that endotoxins from different bacteria, but grown and prepared
under equivalent conditions exhibit different masking susceptibilities.
58
Table 4 Endotoxin recovery of different endotoxins
Recovery of endotoxin from different species over time is shown. Sterile filtrated bacterial supernatants (2.4.1 preparation 1) were diluted to approximately 50 EU/mL in depyrogenated water (positive control). For recovery experiments the endotoxins were spiked with 0.05 wt % polysorbate 20 and 10 mM sodium citrate (reverse mode, 2.5.3.2). Endotoxin recovery was determined after 0, 1, 2, 5 and 7 days of incubation at RT.
Recovery (%)
Source positive control
T0 T1 T2 T5 T7
E.coli O55:B5 40.9 32 7 8 2 0
E.coli O113 64.7 42 0 0 0 0
E.cloacae 50.9 20 4 1 0 0
S.marcescens 60.5 106 46 29 12 8
P.aeruginosa 79.6 135 29 25 23 16
B.cepacia 48.5 248 81 141 92 113
S.maltophilia 37.9 20 1 1 0 0
R.pickettii 68.4 108 73 80 71 55
Furthermore, crude endotoxin extracts from E.coli O113, P.aeruginosa and B.cepacia were
prepared under different growth conditions (2.4.2 preparation 2). Thereby, the bacteria were
grown under conditions including rich-nutrition media (100 % LB) and elevated temperatures (37
°C) as well as low-nutrition (1 % LB) and ambient temperatures (RT).
Figure 19 Comparison of crude endotoxin preparations from different bacteria
Endotoxin recovery is plotted as function of incubation time. 100 EU/mL endotoxin from different bacteria ((A) E.coli O113:H21-, (B) B.cepacia and (C) P.aeruginosa) were spiked into samples containing 10 mM sodium citrate and 0.05 wt % polysorbate 20 and incubated at RT (reverse mode, 2.5.3.2). The used endotoxin extracts were derived from two different bacterial growth conditions (2.4.2 preparation 2). Squares reflect recovery of endotoxin from bacteria grown at 37 °C using 100% LB media. Triangles reflect recovery of endotoxin from bacteria grown at room temperature using 1% (v/v) LB media. Each endotoxin was prepared in triplicate. The corresponding endotoxin measurements of three repetitions were analyzed on the same microtiter plate and for calculation of the data points the mean value of the three individual preparations were used. The error bars represent the standard deviation of the three replicates.
00
50
100
150
200
250
0 5
Re
cove
ry [
%]
A
00
50
100
150
200
250
0 5Time [d]
B
00
50
100
150
200
250
0 5
C
59
Endotoxin from E.coli O113 is not recovered independent of the growth conditions of the bacteria
(Figure 19A) after one day of incubation. In contrast, the recovery of endotoxin prepared from
B.capecia shows no significant decline, regardless of the different growth conditions (Figure 19B).
Interestingly, recovery of endotoxin from P.aeruginosa depends on different growth conditions
(Figure 19C). In this case, modified growth conditions resulted in diverging masking kinetics. To
this end, it has been described that bacterial growth under limitation of divalent cations may also
affect the molecular structures of endotoxin[95]. To further study this effect, bacteria from
E.cloacae were grown under rich nutrition conditions in the presence (EDTA) and absence (TRIS)
of the strong complex forming agent EDTA (2.4.2 preparation 2). The endotoxin recovery kinetics
indicates that both endotoxins are affected by masking over time (Figure 20). However, the
recovery over time of endotoxin, which was prepared out of EDTA-treated bacterial cells (Figure
20B) is greater than the recovery of endotoxin from bacteria which was not treated with EDTA
during bacterial growth (Figure 20A). Comparing the error bars in both experiments, endotoxin
recovery from EDTA treated cells tend to be more variable and by chance, a trend to reduced
recovery over time is not observed. Although chelators have crucial effects on masking of
endotoxin, the presence of chelators during bacterial growth reduces the masking susceptibility
of the endotoxin under given conditions.
Figure 20 Endotoxin recovery kinetics of endotoxin from E.cloacae
Endotoxin recovery is plotted as a function of incubation time. 100 EU/mL endotoxin were spiked into samples containing 10 mM sodium citrate and 0.05 wt % polysorbate 20 and incubated up to seven days at 4°C. The crude endotoxin extracts were derived from E.cloacae and the media (100% LB) for bacterial growth at 37 °C was supplemented with (A) tris buffer and (B) EDTA (2.4.2 preparation 2). Each data point represents the mean of three independent measurements. The slope is obtained by linear fit of the mean data points. The error bars represent the lowest and highest determined values at each time point.
0
50
100
150
200
0 1 4 7
Re
cove
ry [
%]
time [d]
A
0
50
100
150
200
0 1 4 7
Re
cove
ry [
%]
time [d]
B
60
However, for investigation of endotoxin masking in quality control of pharmaceutical industries
standardized endotoxins are requested by regulatory authorities (US FDA). Such standard
endotoxins are prepared by hot-phenol extraction of LPS [91]. Thereby, hydrophobic components
like phospholipids and lipoproteins are removed. To analyze, if masking susceptibilities of
endotoxin are affected by purification, phenol extracted endotoxins from E.coli, S.marcescens and
P.mirabilis were spiked into samples containing a chelator and surfactant and detected over time
(Figure 21). In this case, endotoxin from E.coli was low (14%) in recovery directly after spiking and
no activity of endotoxin was detectable after one day of incubation. The endotoxin from
S.marcescens was detectable (90%) at day 0, but likewise low (21%) after one day of incubation.
Although endotoxin from P.mirabilis decreases over time, significant amounts of endotoxin were
recovered (>43%) at all time points. These results show different masking susceptibilities of the
studied endotoxins, indicating that phenol extraction is not eliminating the heterogeneity of
endotoxins from different bacteria.
Figure 21 Comparison of phenol-extracted endotoxins from different bacteria
Endotoxin recovery is shown over incubation time. 100 EU/mL of three different endotoxins were spiked into samples containing 10 mM sodium citrate and 0.05 wt % polysorbate 20 and incubated up to 4 days at RT. The used endotoxins were phenol-extracted clinical isolates. The black bars reflect endotoxins from E.coli, grey bars reflect P.mirabilis and white bars reflect S.marcescens.
To further evaluate the impact of endotoxin purification, endotoxins from the same bacterial
species (E.coli O55:B5), but prepared by different methods were analyzed. Phenol-extracted
endotoxin and crude supernatants of bacterial suspension were spiked into a polysorbate/citrate
matrix and incubated up to six days at room temperature (Figure 22A). Both preparations show
no recovery after one day of incubation, which confirms the pronounced masking susceptibility
of endotoxin from E.coli. Due to the fast kinetics of masking, the experiment was also performed
at decreased incubation temperature (4 °C) (Figure 22B). The reduced incubation temperature
was chosen, because endotoxin masking can be decelerated, allowing a better resolution of slight
0
50
100
150
200
250
0 1 2 3 4
Re
cove
ry[%
]
time [days]
61
differences in masking susceptibilities (3.1). After 2 days of incubation 50% and after 14 days 14%
of the initial endotoxin content can be detected within the crude extract. In comparison, the
recovery of phenol-extracted endotoxin was low after one day of incubation (27%) and no
significant content of endotoxin was detectable after three days of incubation. Under these
circumstances the detectability of crude endotoxin preparations decreases slower compared to
phenol-extracted endotoxin. However, both preparations of endotoxin are affected by masking.
In consequence, endotoxin masking kinetics can be affected by the extraction method of
endotoxin, but the kinetics rather depends on the source of bacteria.
Figure 22 Endotoxin recovery kinetics of different endotoxin preparations
Recovery of endotoxin from E.coli O55:B5 is shown over incubation time. 100 EU/mL of gel-filtrated endotoxin (black bars) and sterile filtrated bacterial suspension (grey bars) were spiked into samples containing 10 mM sodium citrate and 0.05 wt % polysorbate 20. The samples were incubated up to 14 days at (A) RT and (B) 4°C.
In all of the examples above, the source of endotoxin was known and the endotoxin was
consciously added to samples containing surfactants and chelators. In order to examine a real
endotoxin contamination, the detectability of an endogenous contaminated monoclonal
antibody was analyzed. Therefore, a lyophilized antibody was solubilized in four different buffer
systems and an average endotoxin activity of 135 EU/mL and 114 EU/mL was determined before
and after sterile filtration, respectively (Table 5a). The different buffer systems as well as the
filtration had no major effects on the detectability of the endogenous endotoxin contamination
of the antibody.
0
50
100
150
200
0 1 2 3 6
Re
cove
ry [
%]
0
20
40
60
80
100
120
0 1 2 3 4 7 14
Re
cove
ry [
%]
time[d]
A
B
62
Table 5 Detection of an endogenous endotoxin contamination
Activity of an unknown endotoxin contamination under different buffer conditions is shown. 10 mg/mL
monoclonal antibody (MAK33) were solubilized in A) 25 mM sodium citrate, pH 6.5; B) 10 mM sodium
citrate, pH 7.5; C) 160 mM trehalose, 50 mM sodium phosphate pH 6.2 and D) 10 mM sodium
phosphate, pH 7.5). a) Endotoxin content was determined before and after sterile filtration (0.2 µm).
For endotoxin detection EndoZyme® was used. b) After filtration 0.07 and 0.05 wt % polysorbate 80
were added to the samples 1.1 and 2.1, respectively. 0.04 and 0.05 wt % polysorbate 20 were added to
the samples 3.1 and 4.1, respectively. Endotoxin activity was determined immediately after addition of
polysorbate (day 0) and after incubation of three days (day 3) at RT. For endotoxin detection EndoLISA®
was used.
a before filtration after filtration
# Sample: Activity [EU/mL]
Activity [EU/mL]
1.1 MAK + Buffer A 116.9 109.0
2.1 MAK + Buffer B 140.1 118.3
3.1 MAK + Buffer C 128.7 93.5
4.1 MAK + Buffer D 152.8 136.2
Average 134.6 114.3
Known from previous studies, the simultaneous presence of polysorbate and a chelator like
citrate can mask the endotoxin (3.1 and 3.2). Therefore, polysorbate was added to the antibody
solution. Immediately after addition of polysorbate, the endotoxin content was determined and
comparable contents of endotoxin were detected. However, when the antibody was incubated
for three days at room temperature, the detectable amount of endotoxin significantly decreased
in all of the examined samples (Table 5b). This clearly demonstrates that an endogenous
endotoxin contamination, which reflects a real naturally occurring endotoxin, may be masked in
common formulation matrices.
b day 0 day 3
# Sample:
Activity
[EU/mL]
Activity
[EU/mL]
1.2 MAK + Buffer A + Polysorbate 80 98.5 2.8
2.2 MAK + Buffer B + Polysorbate 80 95.6 2.1
3.2 MAK + Buffer C + Polysorabte 20 117.9 5.3
4.2 MAK + Buffer D + Polysorbate 20 131.7 6.8
63
3.3.3 Discussion
In previous studies (3.1 and 3.2), the phenomenon of LER has been studied, using standardized
endotoxins from E.coli. For analytical applications, standards are indispensable to determine
unknown concentrations and to validate a detection method. However, to investigate if the
phenomenon of LER is limited to the use of standardized endotoxins from E.coli, detectability of
endotoxins from different sources was studied in a typical biopharmaceutical drug product matrix
containing polysorbate and sodium citrate. Phenol-extracted endotoxins (Figure 21), crude
endotoxin extracts (Table 4, Figure 19 and 20) and endogenous endotoxin (Table 5) were
incubated into samples containing polysorbate/citrate. These results show that LER is not limited
to standardized endotoxins from E.coli. Endotoxins from different sources and alternative
preparations of endotoxin can be affected as well as standard endotoxin preparations by LER.
Basically, these results clearly demonstrate different masking susceptibilities of different
endotoxins. LPS reflect a complex group of molecules, which possess a common general
architecture[11] and due to the amphoteric and amphiphilic nature of LPS supramolecular
structures are formed[17], [70]. In the case of masking, it is supposed that the presence of
complex forming agents destabilize the salt bridges of divalent cations between LPS, leading to a
reduced rigidity of the aggregate. The additional presence of a surfactant may then change the
initial supramolecular structure and promote the formation of mixed aggregates, thus limiting
the detection of endotoxin. Taking this assumption into account, differences in the molecular
structure of LPS may explain the diverging masking susceptibilities of endotoxins from different
sources.
3.3.3.1 Molecular heterogeneity of LPS
In general, LPS are very heterogeneous molecules. Already the application of a single source LPS
on a SDS-PAGE results in a ladder of bands (Figure 18). This reflects the nature and number of
sugars within a unit, the nature of the linkages of the sugars as well as the number of repetitive
units. O-antigen sugars appear to be most variable, core structures appear to be less variable and
in turn Lipid A structures are considered as the most conserved part of LPS within a genus.[9],
[96], [97] With regard to the previously described studies on endotoxin masking, it has been
demonstrated that destabilization of the LPS aggregates is the crucial step in masking (3.2). Due
to the fact that the O-antigen only marginally contributes to the ionic stabilization, the
heterogeneity of the O-antigen will be neglected in the following examination. Although lipid A
and core region are supposed to be the most conserved part of LPS, diverse molecular structures
of LPS have been observed[11], [23], [98]. Relevant modifications are expected within the charged
substitutes (e.g. phosphates, amines) as well as acylation (e.g. number, length, saturation) in the
core region and lipid A of LPS.
64
In consequence, LPS from different bacteria may exhibit different molecular structures and
therefore, equivalent prepared endotoxins (phenol-extracted) from different bacteria are
expected to show different kinetics in endotoxin recovery studies. For instance, recovery of
endotoxin from E.coli and S.marcescens is low (< 25 %) already after one day of incubation (Figure
21). In contrast, the endotoxin from P.mirabilis showed substantial detectability over time.
Interestingly, the lipid A structures of E.coli and S.marcescens are assumed to be similar[99],
whereas the proposed molecular structure of lipid A from P.mirabilis exhibits some differences.
Differences are seen for example in the acylation and the substitution of the ester bound
phosphate groups linked to the glucomsamine backbone[100]. Thus, modifications in the lipid A
of P.mirabilis might be a reason for the diverging masking kinetics compared to E.coli and
S.marcescens. A similar behavior was observed when crude preparations of different endotoxins
were compared in recovery kinetics (Figure 19, Table 4). The results show endotoxins which are
rapidly affected by LER (e.g. E.coli O113) and endotoxins which are less susceptible (e.g.
P.aeruginosa). Endotoxin from B.cepacia was not affected by masking within the given conditions.
For B.cepacia, it is also described that the LPS possess an unusual structure. The bacteria lower
the anionic charge of the cell surface by the substitution of 4-amino-4-deoxyarabinose (Ara4N)
residues bound to phosphates of the lipid A backbone[101], [102]. Interestingly, endotoxins from
P.mirabilis and B.cepacia are less susceptible to endotoxin masking and both are known for their
almost stoichiometric substitution of Ara4N[103]. This might explain the limited masking
susceptibility.
Figure 23 Structural modifications of lipopolysaccharides
Depending on the environmental conditions during bacterial growth phosphorylation or acylation can change the molecular LPS structure. Changed LPS structures can result in different interactions between LPS molecules: Ionic interactions between LPS molecules in the presence (A) and absence (B) of divalent cations. (modified figure from Yan et al.[95])
As described above, chelators destabilize ionic interactions between negative charged LPS
molecules and divalent cations. Here, if LPS molecules possess additional positively charged
substituents, like Ara4N, LPS molecules can form ionic interactions without divalent cations to
65
stabilize their supramolecular structures. Under such conditions, LPS structures are independent
from divalent cations and chelation of divalent cations has no effects on stability of the
supramolecular structures of LPS (Figure 23). Consequently, the intercalation of surfactants into
LPS aggregates is limited and the endotoxin is less susceptible to masking under such conditions.
Moreover, different bacteria need different conditions for an optimal growth. Comparing
absorbance of bacterial suspensions from different bacteria and the corresponding detectable
endotoxin activities, huge differences in growth and endotoxin content are observed (Table 3).
However, bacteria are able to adapt themselves to an unfavorable environment to ensure
viability. It is known that bacteria are able to modify their primary LPS structure under certain
growth conditions, in order to reinforce the external membrane to assure best protection against
the environment[95]. Moreover, bacteria possess the ability to alter or regulate their lipid A form
under specific environmental conditions[98]. For instance, after growth of bacteria under divalent
cation limitation (e.g. in the presence of EDTA), their LPS exhibits raised contents of aminoethanol
and Ara4N (Figure 23)[95]. Therefore, using endotoxin from bacteria grown under divalent cation
limitation, displays attenuated recovery kinetics (Figure 20), supporting previous assumption.
However, it has to be pointed out that not all endotoxins are similarly affected, if growth
conditions are modified. Comparing the recovery kinetics of endotoxin from the same source, but
grown under different conditions does not automatically result in a change of the masking
susceptibility (Figure 19). Thus, due to the unknown source of a potential contamination, it is
impossible to predict the species of bacteria, its modifications due to the growth conditions, its
molecular structure and consequently its susceptibility to endotoxin masking.
3.3.3.2 Breakdown products of Gram-negative bacteria
In case of a bacterial endotoxin contamination event, LPS might be present in diverse assemblies.
For example, if viable bacterial cells are present, LPS are embedded in the outer membrane of
the cell to form its outer layer. In addition, LPS can be exposed in so called outer membrane
vesicles (OMV), which are segregated by intact cells to improve their protection. Furthermore,
during cell division or cell death monomers and multimers of LPS can be released from the
bacterial cell. Hence, a set of LPS assemblies (Figure24) can exist in parallel. Noteworthy, the
composition of such assemblies can be diverse. A mixture of LPS and accompanying molecules
like lipoproteins, phospholipids are not necessarily evenly distributed. For example, OMV
frequently contain a high ratio of LPS with extended O-antigens but contain less protein compared
to the originated bacteria[104]. Depending on a particular contamination event, the whole set of
assemblies can be present in a sample. This would be the case, if there is an acute bacterial
contamination event. Otherwise, it is also possible that there are only break down products of
the bacteria present in a sample. This can be the case if break down products of the bacteria are
66
transferred into a sample or parts of the contamination are already eliminated during handling
(e.g. chromatography) of the sample.
Figure 24 Origin of LPS – Bacterial cells and their breakdown products
(A) LPS are the major building block of the outer membrane of Gram-negative bacteria. Break down products of the bacteria can be (B) fragments of bacteria, (C) OMVs which are segregated by the cell and (D) monomers.
Considering not only the heterogeneity of LPS, but also accompanying bacterial components (e.g.
lipids, porins or proteins), a contamination can be very diverse. Keeping this in mind, comparable
experiments with regard to LER are only possible using defined endotoxins. Reference or control
standard endotoxins meet such requirements, because these are highly purified suspensions. For
preparation of such standards, bacterial suspensions from e.g. E.coli pass through a set of
purification steps in order to meet the ordinary requirements for a qualified standard[4]. During
such purification steps, accompanying components like lipids or lipoproteins are removed. The
removed components do not directly contribute to the endotoxic potential, as the Lipid A of LPS
has been identified responsible for toxicity of Gram-negative bacteria[14], [15]. Moreover, the
direct comparison of crude and purified endotoxin preparations gave no indication that the
purification process is responsible for the masking susceptibility of an endotoxin (Figure 22). Only
at reduced incubation temperatures, which decelerate the process of masking, the crude
endotoxin preparations show slower masking kinetics than the highly purified endotoxin. It can
be speculated that the supramolecular structures of the less purified endotoxin are partly
stabilized by its accompanying membrane molecules, but masking is not prevented. Obviously,
crude endotoxin preparations showed also diverse masking susceptibilities (Table 4 and Figure
19) and the recovery of an endogenous contamination was also low over time (Table 5). This
proves that endotoxin masking is not driven by a certain preparation of endotoxin. It can be rather
supposed that the molecular structure of LPS determines if an endotoxin is susceptible to masking
or not. Accompanying molecules other than LPS can modulate the stability against endotoxin
masking, but they cannot prevent it.
3.3.3.3 BET and their standard endotoxins
The different behaviors of endotoxins from different sources raise the question, whether the
established standard endotoxins from E.coli are still adequate in BET? This question was already
67
discussed in the 1970th, at the time when E.coli O113 was determined as source for reference
standard endotoxins (RSE)[24]. Finally, endotoxins from E.coli were chosen, because they were
very well characterized and their toxic effects were studied also in man[28]. The establishment of
endotoxin standards from other species was and is possible, but a likewise deep characterization
would be necessary. Purification is also necessary in order to standardize and enable
comparability of the endotoxin and fulfil the general requirements of a standard. The alternative
use of crude endotoxin extracts like the supernatant of a bacterial suspension might reflect in
certain cases a potential contamination more realistic, but it is very difficult to standardize such
preparations. However, due to the heterogeneity of endotoxins from different sources, it can be
supposed that there will be no single standard available, which reflects the diverse nature of
bacterial endotoxins. The origin of LPS is inevitable connected to the bacteria and in turn an
intrinsic heterogeneity is included. Moreover, it is impossible to predict the source and way of a
bacterial endotoxin contamination in a sample. As a consequence, the masking susceptibility of a
potential contamination is unknown. In order to ensure reliable detection of endotoxin the
masking capability of a sample has to be evaluated. To analyze the masking capability of a sample,
endotoxin recovery studies have to be performed with endotoxin spikes, which are susceptible
to masking. The results above have shown that standardized endotoxins from E.coli exhibit a
pronounced susceptibility to endotoxin masking and represent an appropriate source for
endotoxin recovery studies. However, an endotoxin spike should reflect the worst case with
regard to its masking susceptibility. Although standard endotoxins are susceptible to masking in
the investigated cases, it is not proven that standard endotoxins always reflect a worst case
situation. With regard to the heterogeneity of endotoxins and the diversity of sample
compositions a panel of different endotoxins might be the safest way to determine the masking
capability of a sample and to ensure detectability of a potential contamination. Finally, if a sample
is identified with the capability of endotoxin masking, a suitable detection method has to be
developed in order to detect endotoxin and avoid underestimation of a potential contamination.
Such developments are discussed in the following chapter (3.4).
68
3.4 Demasking of Endotoxin
3.4.1 Introduction
Endotoxin is well detectable in aqueous solutions. However, it becomes significantly less active
(i.e. undetectable) by common detection systems, if it is masked by surfactants and chelators (3.1,
3.2 and 3.3). This may have various consequences. The endotoxin can lose its activity, meaning
that potential endotoxin contaminations in a drug product are basically not harmful anymore
because the endotoxin is masked and pyrogenic reactions are prohibited. For a sustainable
suppression of pyrogenic reactions, masking must be irreversible. It has to be ensured that the
endotoxin will not be demasked in-vivo and becomes pyrogenic again. Noteworthy, modified
endotoxins with significantly reduced pyrogenicity are well known and used as an adjuvant to
enhance efficacy of vaccination[105]. Due to this fact, it cannot be excluded that the masked
endotoxin retains its stimulating effects on the innate immune system, even if the endotoxin is
depyrogenated. Last but not least, the detection of endotoxin in a sample, independent of the
toxicity of present endotoxins, gives an indication about the quality of the tested sample. Thus,
existing test procedures have to be optimized to detect masked endotoxin. Due to the widespread
use of Limulus-based detection methods, a sample-treatment prior to the use of such
conventional test methods is desired. Importantly, successful demasking of endotoxin strongly
indicates that endotoxin is not irreversibly deactivated by masking. In consequence, when
demasking is possible in-vitro, it cannot be excluded that demasking in-vivo is also possible.
In order to develop a demasking approach, the nature of endotoxin and the driving forces of
masking need to be understood. Due to the amphiphilic nature of LPS, it tends to aggregate in
aqueous solutions. The basic cause of aggregation is to lower Gibbs free energy, which is
depending on the inner energy and entropy of a system. The latter is predominant in such cases
and driving hydrophobic effects.[18] Thereby, a variety of supramolecular structures can be
formed, which are obviously depending on the particular conditions (e.g. molecular structure,
salinity and polarity). In order to get an idea of a potential supramolecular structure of amphiphilic
molecules, the concept of packing parameter is a helpful tool. Israelachvili introduced a
dimensionless equation describing the packing parameter S, which in turn is depending on the
molecular volume of the hydrophobic moiety, the length of the fully extended hydrophobic
moiety and the cross-sectional areas of the hydrophilic and hydrophobic moiety. Depending on
the value of such a calculated packing parameter, a particular supramolecular structure of the
amphiphilic molecule can be deduced[106].
69
Figure 25 Molecular shape of an amphiphilic molecule determines its supramolecular structure
The molecular relationship of amphiphilic molecules is illustrated in relation to their supramolecular structure. Depending on the molecular shape of an amphiphilic molecule (e.g. cone, cylinder and inverted cone) a corresponding supramolecular structure is formed (e.g. micellar, bilayers and inverted). (Source: [18])
Figure 25 gives some examples of supramolecular structures depending on their molecular shape.
For example, cone shaped amphiphilic molecules, which often contain only a single hydrophobic
tail tend to form spherical micelles (e.g. polysorbates); truncated or cylinder shaped molecules
(e.g. phospholipids) often contain two hydrophobic tails and form preferably bilayers; and
inverted truncated molecules which contain a pronounced hydrophobic portion tend to form
inverted structures (e.g. LPS).[18] This concept does not fully describe the supramolecular
behavior of amphiphilic molecules, as further parameters like the fluidity of acyl chains influence
likewise the aggregation state. However, the concept of packing parameter helps to understand
the formation of supramolecular structures.
Several studies have been performed to explore structure-function relationships of LPS with
regard to its biological activity[9], [11], [18], [23], [107]–[109]. The primary lipid A structure of LPS
was identified to constitute the endotoxic activity[9]. Moreover, due to the amphiphilic nature of
LPS the effects of their supramolecular structures were examined concerning the endotoxic
activity[18], [66], [109]. Brandenburg and co-workers proposed to extend the term “endotoxic
conformation”, which is used to describe the conformation of a single lipid A molecule required
for optimal triggering of biological effects, to “endotoxic supramolecular conformation” which
denotes the particular organization of lipid A aggregates in physiological fluids causing biological
active LPS [110]. Obviously, there is a relationship between endotoxicity and the supramolecular
structure of LPS (Figure 26). Inverted structures possess a higher degree of endotoxicity compared
to lamellar structures.[18] This might be comprehensible, as the hydrophobic part of LPS tends to
be more accessible in solution. In the case of lamellar structures, the hydrophobic part (lipid A) is
70
shielded by the sugar units (core region and O-antigen) in solution and is consequently less toxic.
Taking this knowledge into consideration, demasking can be achieved by changing the sample
environment affecting the aggregation state of endotoxin. Based on this hypothesis sample
treatment for demasking are investigated below.
Figure 26 Relationship between supramolecular structures and endotoxicity
Correlation between supramolecular LPS structure and bioactivity is shown. Depending on the molecular structure of LPS and the environmental conditions, LPS form certain supramolecular structures. The latter in turns affects the activity of LPS, in which the lamellar structures possess less endotoxicity compared to inverted structures. (Source: [18])
71
3.4.2 Results
Various sample compositions are capable to mask endotoxin and render it undetectable (3.1, 3.2
and 3.2). In order to release masked endotoxin out of polysorbate 20 complexes, samples were
treated with alkyl alcohols (C8 to C18) prior to endotoxin measurements (Figure 27).
Figure 27 Demasking of endotoxin using co-surfactants
Endotoxin recovery after demasking using different concentrations of alkyl alcohols with various chain lengths is shown. 100 EU/mL of endotoxin were spiked into samples containing 10 mM sodium citrate and 0.05 wt % polysorbate 20 and incubated at least for 24 hours at RT. For demasking, samples were treated using alkyl alcohols with varying chain length from C8 to C18 (1-octanol (green bars), 1-decanol (blue bars), 1-dodecanol (black bars), 1-tetradecanol (white bars), 1-hexadecanol (grey bars), 1-octadecanol (orange bars)). The concentrations of the alcohols ranged from 0.6 to 40.0 mM. For detection of endotoxin EndoLISA® was used.
This result indicates that sample treatment of masked endotoxin using 1-dodecanol and 1-
tetradecanol enable substantial recovery. Highest recovery was obtained using concentrations of
5 mM 1-dodecanol and 10 mM 1-tetradecanol, respectively. Using alcohols with alkyl chains
lengths below C12 or above C14 resulted in recovery below 10 %. Thus, the alkyl chain length of
the alkyl alcohol is crucial for demasking, but it clearly demonstrates that masked endotoxin can
be rendered detectable again. There is a small range of 1-dodecanol concentrations which
enabled demasking. In order to improve demasking using 1-dodecanol, 10 mg/mL Bovine Serum
Albumin (BSA) was added to the particular 1-dodecanol concentration. BSA was chosen due to its
capability of binding surfactants (Figure 28). Comparing demasking results in presence and
absence of BSA, a consistent increase in the recovery of endotoxin can be observed in the
presence of BSA.
0
10
20
30
40
50
60
70
80
90
100
40,00 20,00 10,00 5,00 2,50 1,25 0,63
Re
cove
ry [
%]
Concentration [mM]
72
Figure 28 Demasking of endotoxin using dodecanol and BSA
Endotoxin recovery after demasking is shown as function of 1-dodecanol concentration. 100 EU/mL of endotoxin were spiked into samples containing 10 mM sodium citrate and 0.05 wt % polysorbate 20 and incubated for at least 24 hours at RT. Black bars reflect sample treatment using only various concentrations of 1-dodecanol and white bars reflect sample treatment using various concentrations of 1-dodecanol and additional 10 mg/mL BSA. For detection of endotoxin EndoLISA® was used.
To investigate whether demasking is due to similar alkyl chain lengths (C12) of polysorbate 20 and
1-dodecanol, endotoxin was also masked in the presence of polysorbate 80, which possesses C18
alkyl chain and octoxynol 9 which in turn possess a tetramethylbutyl-phenyl group. All of these
surfactants induce LER in combination with sodium citrate (3.1). Sample treatment with 1-
dodecanol leads to recovery of 29 % out of polysorbate 80. Out of samples containing octoxynol
9, no demasking effects were achieved. The combination of 1-dodecanol and BSA lead to a full
recovery out of polysorbate 80, but only to limited recovery out of octoxynol 9 samples (Figure
29).
0
20
40
60
80
100
120
140
160
180
200
40 20 10 5 2,5 1,25 0,625
Re
cove
ry[%
]
1-Dodecanol (mM)
73
Figure 29 Demasking of endotoxin out of different sample matrices using dodecanol and BSA
Endotoxin recovery is shown after demasking as function of 1-dodecanol concentration. 100 EU/mL of endotoxin were spiked into samples containing 0.05 wt % (A) polysorbate 80 and (B) octoxynol 9. All samples were buffered using 10 mM sodium citrate. The samples were incubated for at least 24 hours at RT. For demasking, samples were treated with 1-dodecanol (black bars) as well as with 1-dodecanol and 10 mg/mL BSA (white bars). For detection of endotoxin EndoLISA® was used.
These results indicate that there is a relationship between the alkyl chains of the masking and
demasking components, but the combination of 1-dodecanol and BSA enables demasking out of
polysorbate 20 and 80 masking conditions. Demasking out of octoxynol 9 is less effective,
indicating that the LPS-octoxynol 9 complex is stabilized under these conditions (Figure 29B). To
further enhance demasking, additional agents including sodium dodecylsulfate (SDS) and calcium
dichloride were added to the previously used demasking agents.
Figure 30 Demasking of endotoxin out of octoxynol 9 matrices
Endotoxin recovery is shown in dependence of various demasking compositions. 100 EU/mL of endotoxin were spiked into samples containing 10 mM sodium citrate and 0.05 wt % octoxynol 9 and incubated for at least 24 hours at RT. Demasking was performed using the following components: 5 mM 1-dodecanol (D), 10 mg/mL BSA (B), 100 mM calcium dichloride (C) and 0.1 wt % sodium dodecylsulfate (S). In different demasking approaches, combinations of the components were used (D, BD, CBSD, CBS and BSD). For detection of Endotoxin EndoLISA® was used.
0
20
40
60
80
100
40
,00
20
,00
10
,00
5,0
0
2,5
0
1,2
5
0,6
3
Re
cove
ry [
%]
1-Dodecanol [mM]
A
0
20
40
60
80
100
0.0
40
0.0
20
0.0
10
0.0
05
1-Dodecanol [mM]
B
0
50
100
150
200
D DB CBSD CBS BSD
Re
co
ve
ry [
%]
74
The combination of calcium dichloride, BSA, SDS and 1-dodecanol resulted in a substantial
recovery of endotoxin out of samples containing octoxynol 9 (Figure 30). Table 6 gives an
overview of different demasking approaches out of the different surfactant masking conditions.
These results indicate that the demasking approach including calcium dichloride, BSA, SDS and 1-
dodecanol is suitable for demasking of all examined masking conditions.
Table 6 Comparison of different demasking approaches
Endotoxin recovery after demasking out of different masking surfactants using different approaches is shown. In each case 100 EU/mL of endotoxin were spiked in samples containing 10 mM sodium citrate and 0.05 wt % of the corresponding surfactant (polysorbate 20, polysorbate 80 and octoxynol 9). After incubation of at least 24 hours at RT the samples were treated using 5 mM 1-dodecanol (D), 10 mg/mL BSA (B), 100 mM calcium dichloride (C) and 0.1 wt % sodium dodecylsulfate (S). In different demasking approaches, combinations of the components were used (D, BD and CBSD). For detection of Endotoxin EndoLISA® was used.
D BD CBSD
Masking – Surfactant: Recovery [%] Polysorbate 20 78 170 141 Polysorbate 80 28 94 161 Octoxynol 9 0 23 168
In the experiments shown so far, endotoxin demasking was performed with a commercially
available, highly purified endotoxin preparation from E. coli 055:B5. According to previous studies
(3.3), endotoxin from different sources may have different masking susceptibilities, due to
variances in acyl chain length of the lipid A part of LPS, as well as modifications of side chains[3].
Even more, the length of the O-sugar side chains of LPS potentially impacts the demasking
approach. It cannot be excluded that highly purified endotoxin and crude endotoxin extracts
(often called “NOE”) behave different in demasking mechanism. To address this issue and to
exclude the possibility that the demasking approach is specific for the above used LPS from E. coli
055:B5, endotoxins from different bacteria with different structures and purities were masked in
sample matrixes containing polysorbate 20/citrate (Table 7a), polysorbate 80/citrate (Table 7b)
and octoxynol 9/citrate (Table 7c). For subsequent demasking the various approaches using either
1-dodecanol alone, BSA/1-dodecanol or calcium dichloride/BSA/SDS/1-dodecanol were applied.
The results clearly show that the ability to successfully demask endotoxin from various masking
systems is independent of the source and type of endotoxin used. It shows that demasking is a
general technique applicable to various types of endotoxin from various sources, under a variety
of masking conditions.
75
Table 7 Demasking of different endotoxins
Endotoxin recovery before (masking control) and after demasking of endotoxin from different sources and types out of 10 mM sodium citrate and 0.05 wt % (a) polysorbate 20, (b) polysorbate 80 and (c) octoxynol 9 are shown. Approximately 50 EU/mL of the particular endotoxin were spiked into the corresponding sample matrix. The endotoxins were incubated for seven days at RT in the sample matrix. For demasking, the samples were treated using 5 mM 1-dodecanol (D), 10 mg/mL BSA (B), 100 mM calcium dichloride (C) and 0.1 wt % sodium dodecylsulfate (S). In the particular demasking approaches, combinations of the components were used (D, BD and CBSD). For detection of endotoxin EndoLISA® was used.
a) Polysorbate 20 / sodium citrate
Endotoxin: Source: Masking control
Demasking
D BD CBSD
Recovery [%]
K.pneumonia LMU 0 66 128 212 M.morganii LMU 0 81 110 120 Y.enterocolitica LMU 0 63 174 243 S.marcescens LMU 0 128 168 182 N.meningitis LMU 0 9 23 38 A.baumanni LMU 0 0 124 655 E.cloacae Hyglos 0 55 156 187 S.enterica Sigma 0 42 63 76 E.coli K 12 Invivogen 3 78 80 137 P.aeruginosa Sigma 0 14 5 179
b) Polysorbate 80 / sodium citrate Endotoxin: Source: Masking
control Demasking
D BD CBSD
Recovery [%] K.pneumonia LMU 10 22 12 162 M.morganii LMU 6 35 23 48 Y.enterocolitica LMU 0 13 19 236 S.marcescens LMU 4 28 20 80 N.meningitis LMU 0 55 14 161 A.baumanni LMU 8 0 57 918 E.cloacae Hyglos 0 2 26 85 S.enterica Sigma 0 1 11 25 E.coli K 12 Invivogen 0 21 12 234 P.aeruginosa Sigma 0 54 17 78
c) Octoxynol 9 / sodium citrate Endotoxin: Source: Masking
control Demasking
D BD CBSD
Recovery [%] K.pneumonia LMU 0 12 173 353 M.morganii LMU 15 15 39 99 Y.enterocolitica LMU 7 22 168 309 S.marcescens LMU 0 105 199 326 N.meningitis LMU 0 0 11 42 A.baumanni LMU 0 7 337 511
76
E.cloacae Hyglos 24 27 74 183 S.enterica Sigma 1 1 1 90 E.coli K 12 Invivogen 0 18 10 69 P.aeruginosa Sigma 2 85 106 176
To this end, demasking has been demonstrated in diverse surfactant/buffer matrices. These
matrices were chosen, because pharmaceutical industries often have been using such
components for formulation of APIs like proteins (table 1). Further, antibodies constitute
frequently formulated pharmaceutical protein products. Hence, the established demasking
approaches are applied to systems containing surfactant and an antibody buffered in phosphate
and saline. Polysorbate 20 and 80 were chosen as surfactants (Table 8).
Table 8 Demasking of endotoxin out of formulated antibody samples
Endotoxin recovery before and after demasking out of formulated antibody (PAK) samples is shown. 50 EU/ml of endotoxin were spiked into samples containing water, buffer (10 mM sodium phosphate and 50 mM sodium chloride), antibody (10 mg/mL polyclonal antibody) and surfactant (0.05 wt % polysorbate 20 and polysorbate 80). Samples were incubated for three days at RT. For demasking, the sample containing buffer, surfactant and antibody was treated by using calcium dichloride (C), BSA (B), sodium dodecylsulfate (S) and dodecanol (D). For detection of endotoxin EndoLISA® was used.
Masking surfactant: polysorbate 20
polysorbate 80
Sample: C
[mM] B
[mg/ml] S
[%] D
[mM] Recovery
[%]
water - - - - 100 100
buffer - - - - 102 99
buffer + antibody - - - - 31 44
buffer + surfactant - - - - 0 2
buffer + surfactant + antibody - - - - 0 9
buffer + surfactant + antibody - - - 10.0 17 9
buffer + surfactant + antibody - - - 1.0 20 7
buffer + surfactant + antibody - - - 0.1 0 5
buffer + surfactant + antibody - 10 - 10.0 41 11
buffer + surfactant + antibody - 10 - 1.0 3 6
buffer + surfactant + antibody - 10 - 0.1 2 12
buffer + surfactant + antibody 100 10 0.1 10.0 5 3
buffer + surfactant + antibody 100 10 0.1 1.0 16 23
buffer + surfactant + antibody 100 10 0.1 0.1 67 91
The results show that the buffer solution without polysorbate does not mask the endotoxin.
Buffer solutions containing antibody, but no surfactant, resulted in reduced endotoxin recovery
suggesting that already the antibody contributes a masking effect. The endotoxin recovery from
buffer solutions containing polysorbate and antibody are below 10 % when no demasking
77
treatments were performed. Thus, not only the surfactants but also the antibody is capable of
endotoxin masking. Endotoxin recovery after demasking of such samples, simulating a drug
product containing endotoxin, surfactant, buffer and antibody are low using 1-dodecanol alone
(< 10%). Using a combination of BSA and 1-dodecanol allows moderate endotoxin recovery (10
to 40 %), but a combination of calcium dichloride, BSA, SDS and 1-dodecanol shows a substantial
endotoxin recovery in the presence of Polysorbate 20 and 80 (> 60 %).
To show that demasking is not only possible from solutions containing LPS of a known source, a
commercially available mouse monoclonal antibody for diagnostic purpose was used, which
contained an “endogenous” LPS contamination from an unknown source (Table 9).
Table 9 Demasking of unknown endotoxin
Endotoxin recovery before and after demasking of endotoxin from an unknown source is shown. A contaminated monoclonal antibody (MAK 33) was dissolved in a buffer containing 25 mM sodium citrate (pH 6.5) and 150 mM sodium chloride. Directly after solubilization of the antibody, an endotoxin content of 11 EU/mg was determined. Endotoxin masking was initiated by addition of 0.07 wt % of polysorbate 80 and incubated for three days at RT. For demasking, the sample containing buffer, polysorbate and antibody was treated by using the indicated concentrations of calcium dichloride (C), BSA (B), sodium dodecylsulfate (S) and dodecanol (D). For detection of endotoxin EndoLISA® was used.
Sample: C
[mM] B
[mg/ml] S
[%] D
[mM] Recovery
[%]
buffer + antibody (0 days) - - - - 100
buffer + antibody (3 days) - - - - 57
buffer + polysorbate 80 - - - - 0
buffer + polysorbate 80 + antibody - - - - 3
buffer + polysorbate 80 + antibody - - - 10 45
buffer + polysorbate 80 + antibody - 10 - 10 68
buffer + polysorbate 80 + antibody 100 10 0.1 0.1 178
This antibody was dissolved in a buffer composition corresponding to the formulation of the
known antibody drug product Rituximab containing sodium citrate, sodium chloride and
polysorbate 80 (MabThera®, Rituxan®). The buffer solution containing antibody without
polysorbate masks approximately 40 % of the endotoxin contamination within 3 days of
incubation at room temperature. Incubation in buffer containing either polysorbate 80 or
antibody and polysorbate 80, results in endotoxin recovery below 4%. This shows that an
endogenous endotoxin contamination can be masked and that the risk of masking applies not
only for purified or crude endotoxin extracts, but also for endogenous endotoxin. Demasking of
this endotoxin contamination out of the antibody/surfactant sample resulted in an endotoxin
recovery of 45 % using 1-dodecanol, 68 % using a combination of BSA and 1-dodecanol and 179
% using a combination of calcium dichloride, BSA, SDS and 1-dodecanol. This demonstrates that
the developed approaches are able to demask endotoxin under conditions of relevance for the
78
pharmaceutical industry.For detection of demasked endotoxin, the EndoLISA assay was used in
all experiments shown above. EndoLISA is the method of choice due to its heterogeneous test
format, which reduces test interferences substantially[42]. However, in order to investigate if
other test formats and methods are also applicable after demasking, a recombinant Factor C test
(homogeneous format) as well as a kinetic chromogenic LAL assay was used for detection of
demasked endotoxin. Endotoxin recovery before and after demasking was analyzed out of
polysorbate 20 and 80 in phosphate-buffered saline (PBS) (Table 10). The masking controls
showed no endotoxin recovery in either sample. However, after demasking substantial contents
of endotoxin were recovered in all samples using LAL as well as rFC test methods. This experiment
proves that the detection of demasked endotoxin is independent from the detection system used.
LAL and rFC are suitable test methods after demasking.
Table 10 Comparison of different detection methods after demasking of endotoxin
Endotoxin recovery out of PBS containing 0.05 wt % polysorbate 20 (P20) and polysorbate 80 (P80) is shown, respectively. For masking, approximately 10 EU/mL endotoxin were spiked into the samples and incubated for three days at room temperature. Afterwards, the samples were treated using 200 mM sodium citrate, 100 mM calcium dichloride, 1 mg/mL BSA, 0.1 wt % sodium dodecylsulfate and 0.1 mM 1-dodecanol. For detection of endotoxin rFC (EndoZyme®) and LAL (Kinetic-QCLTM) assays were used.
Recombinant Factor C Limulus Amebocyte Lystate Sample: PBS + P80 PBS + P20 PBS + P80 PBS + P20 [EU/mL] Positive control 9 7 12 7 Recovery [%] Before demasking 0 0 3 0 After demasking 65 66 96 47
79
3.4.3 Discussion
3.4.3.1 Concept of Demasking
Endotoxin masking by surfactants is currently the most prominent masking cause in quality
control of biopharmaceutical drug product manufacturing. During masking, the aggregation state
of LPS is changed, leading to LPS disaggregation and embedment of LPS in surfactant micelles and
in turn the endotoxin becomes undetectable. Considering the molecular shapes of surfactants
like polysorbates (cone shape) and LPS (cylindrical / inverted truncated cone shape), mixed
aggregates result most likely in micellar structures (Figure 31), given that surfactants are in molar
excess. In the case of a potential contamination event, the commonly used surfactants in drug
products are in a great molar excess compared to expected LPS concentrations (3.2).
Figure 31 Potential effects on supramolecular structures: Mixing polysorbate and LPS
Formation of mixed polysorbate-LPS aggregates is shown. (A) Polysorbates possess cone shaped structures and form spherical supramolecular structures. (B) LPS possess cylindrical as well inverted truncated cone shaped structures and form bilayers and inverted supramolecular structures. (C) Mixing polysorbates and LPS, the formation of mixed micelles is predicted, under condition that polysorbates are in molar excess.
To detect masked LPS, LPS have to be liberated from their masking complex. Therefore, it was
searched for molecules, which are capable to destabilize the LPS-surfactant complex and in turn
enable a reassembly of LPS. It is hypothesized that a reorganization of LPS is possible, when
surfactants and LPS do not favor the spherical micellar aggregation state. Concurrently, the
surfactant LPS complex is unfavored and LPS are released and reassembled. Noteworthy, the pure
sample dilution below the critical micelle concentration (CMC) of surfactants was not sufficient
(3.1), assuming that mixtures of LPS and surfactants form stable aggregates. Therefore, the
application of co-surfactants was considered, because the co-surfactants are capable in affecting
the supramolecular arrangement of surfactants (Figure 32).
80
Figure 32 Mixing surfactants and co-surfactants
The change of supramolecular structures is shown when surfactants and co-surfactants are mixed. (A) Cone shaped surfactants (e.g. polysorbate) form spherical micelles above critical micelle concentrations (CMC). (B) Addition of co-surfactant (e.g. dodecanol) to surfactants. (C) Mixing of surfactants and co-surfactants results in alteration of supramolecular structures. The overall hydrophobic portion increases, whereby the hydrophilic portion remains constant. As a result, potentially cylindrical micelles or even bilayers are formed.
Typically, co-surfactants are not able to form micelles because their solubility in water is lower
than their critical micelle concentration[111]. But co-surfactants can intercalate into surfactant
micelles and swell them[112]. Moreover, co-surfactants like long chained alkyl alcohols can
change the overall packing of aggregates and lead to altered aggregation states[113]. In the event
of demasking it is expected that co-surfactants intercalate into both surfactant micelles and mixed
LPS-surfactant micelles. This disturbs the aggregation states of pure and mixed micelles and new
aggregation states are established. It is probable that the surfactants in presence of co-
surfactants no longer prefer the formation of spherical micelles, but rather prefer the formation
of cylindrical and bilayered aggregates. This reorganization may in turn enable the release of LPS
out of masking complex (mixed LPS-surfactant micelles). Moreover, co-surfactants are also
capable to interact with the fatty acids of LPS, which can affect the aggregation state of LPS. It is
supposed that co-surfactants support the reaggregation of LPS and “catalyze” the formation of
inverted truncated cone shaped LPS, which in turn favors the assembly of inverted cubic or
hexagonal supramolecular LPS structures (Figure 33). This is the working hypothesis of demasking
and the application of this concept will be discussed below (3.4.3.2).
81
Figure 33 Potential effects on supramolecular structures: Reassembly of LPS
Potential reassembly of LPS is shown. (A) Mixed LPS-polysorbate aggregates form spherical or cylindrical micelles. The addition of a demasking agent (e.g. co-surfactant) changes the overall packing of aggregates and leads to segregation of surfactants and LPS. (B) Surfactants form cylindrical micelles or bilayers, which no longer stabilize monomeric LPS embedded in micelles. (C) LPS forms preferably hexagonal inverted structures which are well detectable.
3.4.3.2 Realization of demasking concept
According to the above described concept (3.4.3.1), co-surfactants were studied in order to
demask endotoxin. To this end, long-chained alkyl alcohols were used for endotoxin demasking
out of polysorbate 20/citrate samples (Figure 27)[114]. Using alcohols with chain length of C12
resulted in full recovery of endotoxin. The use of C14 alcohols also resulted in a substantial
endotoxin recovery, whereas the use of alcohols with longer or shorter alkyl chains was not
suitable. The beneficial effects of C12 alkyl alcohol might be explained by its chain length, which
fits well to the alkyl chain length of polysorbate 20. Using the C12 alkyl alcohol for demasking out
of a matrix, containing polysorbate 80 (Figure 29Aa), the demasking efficiency is lower. In this
case, the C12 alkyl chain of the alcohol can interact with the C18 of polysorbate as well, but it
possesses a relatively shorter hydrophobic proportion. However, the principle of swelling micelles
is still given. The use of alcohols with longer alkyl chains than C12 or C14 would have been
beneficial for demasking out of polysorbate 80, but was difficult in handling due to their limited
solubility in water. The working solutions of the alkyl alcohols were already solubilized in ethanol
to enhance solubility in aqueous solutions. Nevertheless, by adding the working solutions (alkyl
chain length > C14) to the particular aqueous samples, the ethanol content is likewise diluted and
phase separation occurs. Under such conditions the endotoxin measurement is not necessarily
reliable. It is also not possible to increase the ethanol content in the sample to be tested to
provide better solubility of the long chained alcohols, because of subsequent interference of
ethanol with the enzymatic reaction of the endotoxin assay.
82
Due to the limited applicability of co-surfactants with alkyls chain lengths above C14, the
demasking efficiency out of polysorbate 80 was increased by addition of an ancillary component
capable of binding surfactants. It is supposed that the demasking efficiency of 1-dodecanol is
extended, by limiting the concentration of “free” surfactants in solution. To this end, BSA was
chosen, because it is widely available, well characterized and known to adsorb polysorbates[115],
[116]. The use of 1-dodecanol in combination with BSA resulted in significant enhancement of
endotoxin demasking out of samples in which polysorbate caused endotoxin masking (Figures 28
and 29A). However, in the case of octoxynol 9 masking, using BSA and 1-dodecanol for demasking
was not sufficient. This observation suggests that the masking LPS-octoxynol 9 complex is more
stable than the LPS-polysorbate complex and 1-dodecanol is less effective in this case. Comparing
polysorbates and octoxynol 9, dissimilarities in their nature are given. Polysorbates comprise of
sorbitan substituted with approximately 20 repeat units of polyethylene glycol and an
unbranched alkyl chain. Octoxynol 9 comprises of approximately 10 repeat units of polyethylene
glycol and phenyl with a branched alkyl chain (Figure 34).
Figure 34 Molecular structures of surfactants
Chemical structures of (A) polysorbate 20, (B) polysorbate 80 and (C) octoxynol 9 are displayed.
Polysorbates contain approximately 20 repeat units of ethylenglycol (=w+x+y+z), which are distributed
across four chains. Polysorbate 20 possess a saturated alkyl chain of 12 C-atoms and polysorbate 80
possess an unsaturated alkyl chain of 18 C-atoms. Octoxynol contains 9-10 repeat units of ethylenglycol
(=n), which are connected to a tetramethylbutylphenyl group. (source: [117]–[119])
Comparing hydrophilic hydrophobic balances (HLB) of these surfactants, octoxynol 9 (HLB 13.5)
is more hydrophobic than polysorbate 80 (HLB 15.0) and polysorbate 20 (HLB 16.7)[120].
Furthermore, octoxynol 9 micelles are described to be more asymmetric than polysorbate
micelles and bind less water than those composed of polysorbate[121]. Hence, it is conceivable
that octoxynol 9 forms more stable aggregates with LPS compared to polysorbates due to its more
pronounced hydrophobic nature. In order to destabilize the LPS-octoxynol 9 complex, it was
searched for a well characterized charged surfactant, because these are harsher than non-ionic
surfactants[122]. To this end, SDS was chosen, because of its anionic nature, which does not favor
ionic interactions with LPS and has the capability to mix very well with octoxynol 9[123].
83
Moreover, SDS has a high affinity to bind to proteins[124]. This effect may also be beneficial for
displacing LPS adsorbed to BSA and further proteins like antibodies. Taking these considerations
into account, a combination of SDS, BSA and 1-dodecanol was used for demasking. However, the
full content of endotoxin could not be retrieved, so that further optimization was needed. It was
supposed that the masking complex is still too rigid under given conditions. It has been described
that an increase of ionic strength in octoxynol 9/SDS mixtures can significantly change the
aggregation state[125]. Especially in the presence of calcium dichloride, SDS aggregates are
swollen and progress from prolate ellipsoids to extended cylinders or rods[126]. Furthermore,
divalent cations stabilize LPS aggregates, which may also support the reassembly of LPS and
neutralize chelators. Using the combination of calcium dichloride, BSA, SDS and 1-dodecanol
(CBSD) for demasking, endotoxin was successfully detected, when it was masked in the presence
of octoxynol 9 (Figure 30). These results show that depending on the masking condition different
approaches are necessary to demask endotoxin. Interestingly, the most complex approach (CBSD)
was also suitable for demasking out of polysorbates and octoxynol.
3.4.3.3 Demasking of endotoxins from different sources
To challenge the described demasking approach, endotoxins from diverse sources were masked
using different surfactants and treated with different approaches. Summarizing these results,
sample treatment using CBSD enabled demasking of all endotoxins out of all masking conditions.
The approach using only 1-dodecanol was expected to be sufficient for demasking out of
polysorbate 20 (Table 7a). In fact, most of the endotoxins could be adequately recovered, but a
few endotoxins were limited in recovery. This might be explained by the fact that LPS from
different bacteria as well as originated from various conditions may possess different molecular
structures. There are differences in the length and number as well as modifications in linearity
and saturation of the acyl chains in the lipid A part of LPS. Moreover, variations in the
composition, decoration and length of the sugar residues in the core region and O-antigen are
given[3], [98]. These molecular modifications can have effects on the nature and assembly of LPS.
As assumed before, endotoxin masking is driven by hydrophobic effects, and formation of mixed
aggregates with additional amphiphilic molecules. Thus, the stability of a LPS masking complex is
not necessarily only depending on the masking components, but also on the molecular structure
of LPS. Consequently, to demask some of the endotoxins (e.g. Acinetobacter baumanni), the more
complex and harsh approach (CBSD instead of D) was necessary for full recovery of endotoxin.
Moreover, it can be noticed that the detected activity is sometimes partly enhanced compared
to the detected activity in pure water of the particular endotoxin (Table 7). Basically, the
molecular structure of endotoxin defines the potential activity of endotoxin[11], but the
formation of a certain supramolecular structure modulates the detectable activity[18], which is
84
also depending on the environmental conditions[127]. Thus, inverted structures are more active
than lamellar structures[16], [18]. This is comprehensible, because the activity depends on the
interaction of a receptor (e.g. Factor C) and lipid A of LPS, which in turn is better accessible by
inverted than by regular supramolecular structures. For this reason, it is also possible that LPS in
water possess not exactly the same supramolecular structure compared to LPS after demasking,
resulting in a diverging detectable activity. However, the overall results in recovery of different
endotoxins after demasking are in a passable range, considering variabilities in Limulus-based
detection methods and the heterogeneity of LPS.
3.4.3.4 Influence of proteins on demasking
In order to simulate the impact of a protein-based API on demasking, samples containing
formulated polyclonal and monoclonal antibodies were studied. Also in these cases, the full
contents of endotoxin could be recovered after demasking. The applied approaches were
similarly effective as in the absence of a protein during masking. Nevertheless, it is observed that
a significantly reduced concentration of 1-dodecanol is sufficient for successful demasking in the
presence of an antibody (table 8). Further demasking studies have also shown that the required
concentrations of demasking components can vary, depending on the concentration and
composition of the analyzed sample (data not shown). In consequence, using the discovered
components for demasking, the required concentrations of demasking components have to be
adjusted individually. Although there is a good perception of the masking and demasking
principles, it is difficult to predict the interplay and aggregation state of all sample components
while endotoxin masking and demasking. Hence, the described molecules used for sample
preparation represent a toolbox of demasking agents. To develop a dedicated sample preparation
protocol for demasking endotoxin, broad approaches with various combinations and
concentrations of all described demasking components are recommended.
In the case of biopharmaceutical drug products, the API (mainly proteins) is the major component,
which predetermines the overall sample conditions. Surfactants are added to such protein based
products to saturate hydrophobic interfaces in order to prevent adsorption and aggregation of
the API[56]. The results above have been shown that that the phenomenon of LER can be driven
by the formulation components lacking the API (3.1 and 3.2), but the API can also contribute to
masking of endotoxin and initiate LER (Table 8 and 9). In contrast, BSA has been shown being
beneficial during demasking. To this end, depending on the composition and aggregation state of
sample compositions, more or less surfactants can be adsorbed by a protein[116] and
noteworthy, proteins are also capable in adsorbing LPS[47]. Thus, the presence of a protein can
enhance or reduce the endotoxin masking capability of a sample. It is assumed that the “free”
85
concentration and aggregation state of surfactants in a sample can be affected by the API, which
in turn can influence masking and demasking of endotoxins.
3.4.3.5 Endotoxin demasking – rearrangement of endotoxin aggregates
In summary, depending on the particular sample conditions a combination of 1-dodecanol, SDS,
BSA and calcium dichloride can render the masked endotoxin detectable again. Noteworthy, 1-
dodecanol represents the essential reagent and was present in all demasking approaches. It is
supposed that 1-dodecanol provides the major driving force in disturbing endotoxin masking
complexes and supporting the rearrangement of LPS. Figure 35 schematically illustrates
hypothetical rearrangements of lipid A and polysorbate 20 in the presence of a long-chained alkyl
alcohol. The illustration emphasizes the pass through several transition states, in which co-
surfactants force swelling of the endotoxin masking complex, followed by forming lamellar
structures, which in turn enable a reassembly of detectable LPS.
Figure 35 Potential effects on supramolecular structures: Re-arrangements during demasking
Hypothetical rearrangements of endotoxin during demasking are shown. (A) The lipid A part of LPS is embedded in the hydrophobic core of a small sized surfactant micelle. (B) Co-surfactants intercalate into the mixed surfactant-LPS micelle and swell it. (C) Intercalation of co-surfactants rearranges the micellar structures into lamellar and channel structures. (D) LPS molecules are free to diffuse. (E) Surfactants form bilayers and LPS reassembles into detectable structures (e.g. hexagonal inverted).
In conclusion, the data demonstrate that the detection of endotoxin depends on the particular
sample conditions. LPS are reversibly deactivated during masking, because LPS can be detected
again after sample treatment. Concomitantly, it cannot be excluded that demasking also takes
place in-vivo. Thus, masking of endotoxin is most likely driven by alterations in the supramolecular
structures of endotoxin, which is controlled by environmental conditions.
86
4 Conclusions
4.1 Endotoxin demasking – a technical solution LER has been observed during quality control of biopharmaceutical drug products using Limulus-
based detection systems[44], [54]. As consequence, the detection of bacterial endotoxins can
lead to wrong-negative test results. In this work, the detectability of endotoxins in typical
formulation matrices of biopharmaceutical drug products was analyzed in order to understand
and overcome LER. The results demonstrate that LER is caused by simultaneous presence of
surfactants and complex forming buffer agents. The appearance of LER is time and temperature
dependent and complex forming agents were identified in limiting the reaction kinetics. Variation
in surfactant and endotoxin concentrations showed no substantial effects on the reaction
kinetics, but endotoxins from different sources showed effects on the kinetics. Taken together,
the results above lead to the assumption that LER is caused by alteration of the endotoxin
aggregation state. Moreover, a two-step masking mechanism is proposed, in which salt bridges
between LPS molecules are destabilized and subsequently mixed micelles are formed masking
the endotoxin. In order to render endotoxin detectable again, a sample treatment procedure was
developed. Thereby, dodecanol was identified very efficient in demasking the endotoxin. The
presented results clearly demonstrate that demasking is possible out of various formulation
matrices and independent of the endotoxin source. However, the experiments were based on
model systems and the conditions of drug products were simulated. The ultimate proof of
concept is the application in a real drug product which is affected by LER. The cooperation with a
world´s leading pharmaceutical company enabled the analysis of endotoxin masking in one of
their biopharmaceutical drug products, which is intended for commercial use. The studied
product was unequivocally affected by LER (Figure 36A).
Figure 36 Recovery endotoxin before and after demasking out of a drug product
Endotoxin recovery before and after demasking out of a real life sample is shown. A) Recovery of 10 EU/mL endotoxin after incubation for 7 days at 4 °C in depyrogenated water (control) and finished drug product (DP). B) Recovery of endotoxin out of the drug product after sample treatment. For detection a LAL assay was used. The error bars reflect the standard deviation of four individual sample preparations (n=4).
0
50
100
150
Control DP
Re
cove
ry [
%]
A
0
50
100
150
Recovery referred to control
Rec
ove
ry [
%]
B
87
For demasking the developed toolbox including ionic and amphiphilic demasking agents was used
and a dedicated sample treatment protocol was established to overcome the LER effect in the
drug product. Application of the protocol restored the detectability of endotoxin out of the
affected drug product (Figure 36B). As expected, not all endotoxins show equivalent masking
susceptibilities. For example, endotoxins from E.coli O55:B5 are less affected than E.coli
O113:H21 (RSE) and E.cloacae (NOE). However, after sample treatment, all endotoxins were
detectable in a range between 50 and 200% recovery. Interestingly, the degree of masking had
no impact on demasking efficiency (Figure 37). This approach enabled for the first time an
adequate detection of endotoxin over time in this drug product. Consequently, to ensure the
detection of potential endotoxin contaminations and to reduce the risk of underestimation of an
endotoxin contamination, the sponsor will use this approach in quality control departments in
the future to improve patient safety.
Figure 37 Masking and demasking of different endotoxin out of a drug product
Endotoxin recovery of 2.5 EU/mL of different endotoxins out of a finished drug product after incubation for 15 days at 4°C is shown. Endotoxin was measured before (black bars) and after demasking (white bars). For detection a LAL assay was used.
4.2 Perspectives of sample treatments in BET The occurrence of LER demonstrates that the requirements for endotoxin testing of modern
biological drug products are changing. To this end, snapshot measurements of endotoxin will be
extended by time dependent measurements and trends have to be identified. Moreover, the
complexity of present and future drug products will not decrease and in consequence, it is
expected that individual sample preparations prior to the actual test methods will increase.
Moreover, there are further exciting fields of application, which suffer from inadequate
endotoxin detection. For example, vaccines can exhibit difficult conditions for BET, because of
complex sample formulation including aluminum-based nano particles, which are able to adsorb
endotoxin and strongly interfere with the enzyme reaction in Limulus-based detection systems.
In such a case, the optimization of given detection methods is of great interest. Another challenge
0
50
100
150
200
250
Rec
ove
ry [
%]
88
in BET is the group of Advanced Therapy Medical Products, which possess enormously increased
sample complexity, because such products often contain living cells. However, the ultimate
challenge of endotoxin testing is the analysis of blood samples, because of pronounced masking
effects and strong test interference, which substantially reduces sensitivity of the test system.
Therefore, fast and sensitive detection of endotoxin in blood samples would be a great
achievement in the field of sepsis diagnosis, which in turn would substantially support decisions
in the medical treatment of acute infections. Taken together, the presented data contributes to
a better understanding of endotoxins and helps to improve detection of bacterial endotoxins in
complex sample matrices.
4.3 Need for structural analysis of endotoxins in complex sample
matrices Goal of the present work was to establish a technical solution for endotoxin detection in the case
of endotoxin masking (LER) of biopharmaceutical drug products. To this end, a technical solution
was developed and a mechanistic model was established, assuming structural rearrangements of
LPS during masking and demasking. Yet, to confirm the working hypothesis and to further
improve the current methods, detailed structural analysis of endotoxin will be of interest. It might
be conceivable to track a change of endotoxin aggregates during masking and demasking with
physical methods. For instance, it is most likely that the size of aggregates is changed, which could
be determined using scattering methods like Dynamic Light Scattering (DLS) or microscopy
methods like Atomic Force Microscopy (AFM). The application of such methods is highly
appreciated, but there are a few obstacles that need to be overcome. Due to the heterogeneity
of endotoxin, it exists in a broad variety of aggregates with different shapes and sizes. Thus, the
analysis of simultaneous alterations of different aggregates will be difficult. A further challenge
is the particular endotoxin concentration. Basically, the aggregation state is concentration
dependent and relevant LPS concentrations are in the pico- to femtomolar range, which
challenges the detection limit of most analytical methods. Further difficulties are given by the
molar excess of surfactants compared to LPS. To this end, the detection of structural alterations
of LPS might be interfered by surfactant aggregates. Due to these difficulties, it might be
reasonable to start experiments using the above studied polysorbate/citrate matrix as sample,
but with reduced polysorbate concentrations still allowing for masking, but possibly reducing
interference. In addition, using rough mutants lacking the O-antigen as endotoxin source could
be beneficial in such experiments, because a reduced heterogeneity of endotoxin can be
achieved. Obviously, such conditions do not reflect real conditions in industry, but will support
the understanding of endotoxin aggregation. This further elucidation of supramolecular
89
alterations can be used in diagnostics for improving endotoxin detection methods, but also
support fundamental research of stabilities in bacterial membranes.
4.4 Clinical relevance of masked endotoxin Beyond the presented analytical approach, it is also of great interest to further understand the
clinically effects of masked endotoxin. Endotoxin, when masked could be assumed as
depyrogenated i.e. the endotoxic activity is neutralized and detection could be assumed as
dispensable. At a first glance, this is a solid argument, because Limulus-based detection methods
are useful for identification of LPS and LPS-like structures in samples like biopharmaceutical drug
products. Comparative studies have shown that activities measured with Limulus-based detection
methods indicate bacterial contaminations very sensitive. However, it has to be kept in mind that
Limulus-based detection is derived from an invertebrate crab and is not an in-vivo measure for
endotoxicity in man. Obviously, to study the real pyrogenicity of masked endotoxin, it must be
administrated intravenously to man under a variety of conditions (varying concentrations,
different endotoxin sources, etc.). Yet, such kinds of studies do not correspond with ethic
guidelines. Alternative test procedures are experiments in animals. In Europe it is difficult to
perform such studies, because of animal welfare directives of the European Commission.
Nevertheless, a few unpublished studies using the RPT have been performed, indicating that
rabbits can positively respond to masked endotoxin, but not imperatively. Using another in-vitro
method which mimics interaction of endotoxin and Toll-like receptors of the human innate
immune system might be also beneficial[30]. Interestingly, first results indicate that monocyte
activation tests (MAT) are also affected by LER. The interaction of masked endotoxin with Toll-
like receptor in the assay seems to be not possible. However, the MAT does not reflect in-vivo
conditions. It has to be investigated, whether other physiological functions are needed to break
up the masked endotoxin complex. Otherwise complex formation of endotoxin by chelator and
surfactant might inhibit immune stimulation via Toll-like receptor immune response. In
consequence, there is no definite statement available, if masked endotoxin is still hazardous in
man. Moreover, it has to be considered that masked endotoxin might activate alternative
immune stimulating pathways. In the field of vaccination, depyrogenated endotoxins are actively
added to certain drug products to serve as an adjuvant, stimulating the immune system. Such
effects are desired and support the development of certain adaptive immunity. In contrast,
biopharmaceutical drug products (e.g. therapeutic proteins), which are focus in the present work,
can lose their efficacy or lead to life threatening responses due to innate immune response
modulating impurities.[128], [129] Hence, in the case of therapeutic monoclonal antibodies, the
presence of masked endotoxin could lead to inadvertent side reactions. Consequently,
90
continuous improvement of endotoxin detection methods is essential to maintain and improve
patient safety.
91
4.5 Key findings Endotoxin testing is mandatory in quality control of parenteral drug products. Low
recovery of known endotoxin contents has led to the presented work and resulted in the
following findings:
The occurrence of the Low Endotoxin Recovery (LER) is time and temperature dependent.
LER can be detected after minutes to hours and days of sample incubation depending on
the experimental setup. To thoroughly identify if a sample (e.g. drug product) is affected
by LER, incubation temperature and periods have to reflect handling and storage
procedures of tested samples.
LER is caused by endotoxin masking. Endotoxin detection assays have been proven
functional, but detectability of endotoxin is limited due to alteration of its supramolecular
aggregation state, which in turn can be manipulated by the sample matrix. It is supposed
that endotoxin is monomerized in its masked state.
Common formulation components of biopharmaceutical drug products like surfactants
and buffer systems as well as proteins can lead to LER. However, only the simultaneously
presence of amphiphilic molecules and complex formation agents cause LER. The
complex formation capability of a sample matrix strongly determines the reaction rate.
Endotoxin masking is associated with a two-step reaction mechanism. In a first step,
endotoxin aggregates are permeabilized by destabilization of salt bridges between
endotoxin molecules. In a second step, amphiphilic molecules like surfactants intercalate
between endotoxin molecules and result in masking of endotoxin.
Endotoxins from different sources possess different susceptibilities to endotoxin
masking. Depending on the molecular structure of endotoxin, different stabilization
mechanisms are used. Endotoxins with substantial contents of positively charged
substituent’s (e.g. 4-amino-4-deoxyarabinose) are less susceptible to masking.
Demasking of endotoxin is possible. Endotoxin can be released from its masking complex
and detected again using common detection methods. Sample treatments using
dodecanol, calcium dicholoride, sodium dodecyl sulfate and bovine serum albumin have
been shown to be very effective. It is most likely that endotoxin reaggregates and forms
inverted aggregates.
Sample treatment for demasking is case related. Depending on the sample matrix, above
described components and combination thereof as well as concentrations have to be
adapted case by case. Co-surfactants like long chain alkyl alcohols have been identified
as key components and are necessary in all cases.
92
93
5 Appendix
5.1 List of figures Figure 1 Gram-negative Bacteria .................................................................................................. 13
Figure 2 The history of endotoxin research .................................................................................. 14
Figure 3 Schematic structure of lipopolysaccharides ................................................................... 16
Figure 4 Schematic representation of the activation mechanisms induced by LPS ..................... 16
Figure 5 The Toll-Like-Receptor family ......................................................................................... 17
Figure 6 Horseshoe crabs and their endotoxin specific reaction cascade .................................... 19
Figure 7 Endotoxin recovery kinetics in citrate-polysorbate formulations .................................. 32
Figure 8 Impact of buffer system on endotoxin recovery ............................................................. 34
Figure 9 Surfactant dependent endotoxin recovery ..................................................................... 35
Figure 10 Endotoxin masking capacity of citrate-polysorbate 20 formulation............................. 36
Figure 11 Two-step mechanism of endotoxin masking ................................................................ 39
Figure 12 Endotoxin recovery depending on order of matrix component and LPS addition ....... 42
Figure 13 Endotoxin recovery kinetics depending on LPS concentration ..................................... 43
Figure 14 Endotoxin recovery kinetics depending on concentration of matrix components ...... 44
Figure 15 Endotoxin recovery depends on the presence of divalent cations ............................... 45
Figure 16 Sodium citrate concentration determines endotoxin recovery .................................... 46
Figure 17 Endotoxin recovery depends on sodium citrate concentration under equilibrium
conditions ...................................................................................................................................... 47
Figure 18 SDS-PAGE of crude bacterial endotoxin preparations from different bacteria ............ 57
Figure 19 Comparison of crude endotoxin preparations from different bacteria ........................ 58
Figure 20 Endotoxin recovery kinetics of endotoxin from E.cloacae ............................................ 59
Figure 21 Comparison of phenol-extracted endotoxins from different bacteria ......................... 60
Figure 22 Endotoxin recovery kinetics of different endotoxin preparations ................................ 61
Figure 23 Structural modifications of lipopolysaccharides ........................................................... 64
Figure 24 Origin of LPS – Bacterial cells and their breakdown products ...................................... 66
Figure 25 Molecular shape of an amphiphilic molecule determines its supramolecular
structure ........................................................................................................................................ 69
Figure 26 Relationship between supramolecular structures and endotoxicity ............................ 70
Figure 27 Demasking of endotoxin using co-surfactants .............................................................. 71
Figure 28 Demasking of endotoxin using dodecanol and BSA ...................................................... 72
Figure 29 Demasking of endotoxin out of different sample matrices using dodecanol and BSA . 73
Figure 30 Demasking of endotoxin out of octoxynol 9 matrices .................................................. 73
Figure 31 Potential effects on supramolecular structures: Mixing polysorbate and LPS ............. 79
94
Figure 32 Mixing surfactants and co-surfactants .......................................................................... 80
Figure 33 Potential effects on supramolecular structures: Reassembly of LPS ............................ 81
Figure 34 Molecular structures of surfactants .............................................................................. 82
Figure 35 Potential effects on supramolecular structures: Re-arrangements during demasking 85
Figure 36 Recovery endotoxin before and after demasking out of a drug product ..................... 86
Figure 37 Masking and demasking of different endotoxin out of a drug product ........................ 87
95
5.2 List of tables Table 1 Common formulations of biopharmaceutical drug products .......................................... 21
Table 2 Endotoxin recovery out of single and multiple component samples ............................... 31
Table 3 Growth of different bacteria and release of endotoxin ................................................... 56
Table 4 Endotoxin recovery of different endotoxins .................................................................... 58
Table 5 Detection of an endogenous endotoxin contamination .................................................. 62
Table 6 Comparison of different demasking approaches ............................................................. 74
Table 7 Demasking of different endotoxins .................................................................................. 75
Table 8 Demasking of endotoxin out of formulated antibody samples ....................................... 76
Table 9 Demasking of unknown endotoxin ................................................................................... 77
Table 10 Comparison of different detection methods after demasking of endotoxin ................. 78
96
5.3 References [1] C. R. Woese, „Bacterial evolution“, Microbiol. Rev., Bd. 51, Nr. 2, S. 221–271, Juni 1987.
[2] T. Gregersen, „Rapid method for distinction of gram-negative from gram-positive
bacteria“, Eur. J. Appl. Microbiol. Biotechnol., Bd. 5, Nr. 2, S. 123–127, 1978.
[3] M. Caroff und D. Karibian, „Structure of bacterial lipopolysaccharides“, Carbohydr. Res.,
Bd. 338, Nr. 23, S. 2431–2447, Nov. 2003.
[4] E. T. Rietschel und H. Brade, „Bacterial endotoxins“, Sci. Am., Bd. 267, Nr. 2, S. 54–61, Aug.
1992.
[5] P. O. Magalhães, A. M. Lopes, P. G. Mazzola, C. Rangel-Yagui, T. C. V. Penna, und A. Pessoa,
„Methods of endotoxin removal from biological preparations: a review“, J. Pharm. Pharm.
Sci. Publ. Can. Soc. Pharm. Sci. Société Can. Sci. Pharm., Bd. 10, Nr. 3, S. 388–404, 2007.
[6] H. Brade, Hrsg., Endotoxin in health and disease. New York: Marcel Dekker, 1999.
[7] E. T. Rietschel und J.-M. Cavaillon, „Richard Pfeiffer and Alexandre Besredka: creators of
the concept of endotoxin and anti-endotoxin“, Microbes Infect., Bd. 5, Nr. 15, S. 1407–
1414, Dez. 2003.
[8] R. Pfeiffer, „Untersuchungen über das Choleragift“, Med. Microbiol. Immunol. (Berl.), Bd.
11, Nr. 1, S. 393–412, 1892.
[9] E. T. Rietschel, T. Kirikae, F. U. Schade, U. Mamat, G. Schmidt, H. Loppnow, A. J. Ulmer, U.
Zähringer, U. Seydel, und F. Di Padova, „Bacterial endotoxin: molecular relationships of
structure to activity and function“, FASEB J. Off. Publ. Fed. Am. Soc. Exp. Biol., Bd. 8, Nr. 2,
S. 217–225, Feb. 1994.
[10] O. Holst, A. J. Ulmer, H. Brade, H. D. Flad, und E. T. Rietschel, „Biochemistry and cell biology
of bacterial endotoxins“, FEMS Immunol. Med. Microbiol., Bd. 16, Nr. 2, S. 83–104, Dez.
1996.
[11] C. Erridge, E. Bennett-Guerrero, und I. R. Poxton, „Structure and function of
lipopolysaccharides“, Microbes Infect., Bd. 4, Nr. 8, S. 837–851, Juli 2002.
[12] S. Kadis, G. Weinbaum, und S. J. Ajl, Microbial toxins. Volume IV, Volume IV,. 1971.
[13] P. J. Hitchcock, L. Leive, P. H. Mäkelä, E. T. Rietschel, W. Strittmatter, und D. C. Morrison,
„Lipopolysaccharide nomenclature--past, present, and future“, J. Bacteriol., Bd. 166, Nr. 3,
S. 699–705, Juni 1986.
[14] A. B. Schromm, K. Brandenburg, H. Loppnow, A. P. Moran, M. H. J. Koch, E. T. Rietschel,
und U. Seydel, „Biological activities of lipopolysaccharides are determined by the shape of
their lipid A portion: Bioactivity of LPS determined by lipid A shape“, Eur. J. Biochem., Bd.
267, Nr. 7, S. 2008–2013, Apr. 2000.
97
[15] C. Galanos, O. Lüderitz, E. T. Rietschel, O. Westphal, H. Brade, L. Brade, M. Freudenberg,
U. Schade, M. Imoto, und H. Yoshimura, „Synthetic and natural Escherichia coli free lipid
A express identical endotoxic activities“, Eur. J. Biochem. FEBS, Bd. 148, Nr. 1, S. 1–5, Apr.
1985.
[16] K. Brandenburg, H. Mayer, M. H. Koch, J. Weckesser, E. T. Rietschel, und U. Seydel,
„Influence of the supramolecular structure of free lipid A on its biological activity“, Eur. J.
Biochem. FEBS, Bd. 218, Nr. 2, S. 555–563, Dez. 1993.
[17] K. Brandenburg, J. Andrä, M. Müller, M. H. . Koch, und P. Garidel, „Physicochemical
properties of bacterial glycopolymers in relation to bioactivity“, Carbohydr. Res., Bd. 338,
Nr. 23, S. 2477–2489, Nov. 2003.
[18] P. Garidel, Y. Kaconis, L. Heinbockel, M. Wulf, S. Gerber, A. Munk, V. Vill, und K.
Brandenburg, „Self-Organisation, Thermotropic and Lyotropic Properties of Glycolipids
Related to their Biological Implications“, Open Biochem. J., Bd. 9, Nr. 1, S. 49–72, Sep. 2015.
[19] K. Takayama, Z. Z. Din, P. Mukerjee, P. H. Cooke, und T. N. Kirkland, „Physicochemical
properties of the lipopolysaccharide unit that activates B lymphocytes“, J. Biol. Chem., Bd.
265, Nr. 23, S. 14023–14029, Aug. 1990.
[20] Centers for Disease Control and Prevention (CDC), „Red blood cell transfusions
contaminated with Yersinia enterocolitica--United States, 1991-1996, and initiation of a
national study to detect bacteria-associated transfusion reactions“, MMWR Morb. Mortal.
Wkly. Rep., Bd. 46, Nr. 24, S. 553–555, Juni 1997.
[21] Centers for Disease Control and Prevention (CDC), „Clinical sepsis and death in a newborn
nursery associated with contaminated parenteral medications--Brazil, 1996“, MMWR
Morb. Mortal. Wkly. Rep., Bd. 47, Nr. 29, S. 610–612, Juli 1998.
[22] Centers for Disease Control (CDC), „Postsurgical infections associated with an extrinsically
contaminated intravenous anesthetic agent--California, Illinois, Maine, and Michigan,
1990“, MMWR Morb. Mortal. Wkly. Rep., Bd. 39, Nr. 25, S. 426–427, 433, Juni 1990.
[23] C. R. H. Raetz und C. Whitfield, „Lipopolysaccharide Endotoxins“, Annu. Rev. Biochem., Bd.
71, Nr. 1, S. 635–700, Juni 2002.
[24] K. L. Williams, Hrsg., Endotoxins: pyrogens, LAL testing and depyrogenation, 3rd ed. New
York: Informa Healthcare, 2007.
[25] H. D. Hochstein, „The LAL test versus the rabbit pyrogen test for endotoxin detection:
Update 87“, Pharm Technol, Bd. 11, Nr. 6, S. 124–129, 1987.
[26] R. Medzhitov, „Toll-like receptors and innate immunity“, Nat. Rev. Immunol., Bd. 1, Nr. 2,
S. 135–145, Nov. 2001.
98
[27] W. R. Keene, H. R. Silberman, und M. Landy, „Observations on the pyrogenic response and
its application to the bioassay of endotoxin“, J. Clin. Invest., Bd. 40, Nr. 2, S. 295–301, Feb.
1961.
[28] S. E. Greisman und R. B. Hornick, „Comparative Pyrogenic Reactivity of Rabbit and Man to
Bacterial Endotoxin“, Exp. Biol. Med., Bd. 131, Nr. 4, S. 1154–1158, Sep. 1969.
[29] M. Daneshian, A. Guenther, A. Wendel, T. Hartung, und S. von Aulock, „In vitro pyrogen
test for toxic or immunomodulatory drugs“, J. Immunol. Methods, Bd. 313, Nr. 1–2, S. 169–
175, Juni 2006.
[30] T. Hartung, „The human whole blood pyrogen test - lessons learned in twenty years“,
ALTEX, Bd. 32, Nr. 2, S. 79–100, 2015.
[31] J. Levin und F. B. Bang, „A DESCRIPTION OF CELLULAR COAGULATION IN THE LIMULUS“,
Bull. Johns Hopkins Hosp., Bd. 115, S. 337–345, Okt. 1964.
[32] E. H. Mürer, J. Levin, und R. Holme, „Isolation and studies of the granules of the
amebocytes of Limulus polyphemus, the horseshoe crab“, J. Cell. Physiol., Bd. 86, Nr. 3, S.
533–542, Dez. 1975.
[33] S. Iwanaga, „The limulus clotting reaction“, Curr. Opin. Immunol., Bd. 5, Nr. 1, S. 74–82,
Feb. 1993.
[34] T. Nakamura, T. Morita, und S. Iwanaga, „Lipopolysaccharide-sensitive serine-protease
zymogen (factor C) found in Limulus hemocytes. Isolation and characterization“, Eur. J.
Biochem., Bd. 154, Nr. 3, S. 511–521, Feb. 1986.
[35] J. H. Jorgensen und R. F. Smith, „Preparation, sensitivity, and specificity of Limulus lysate
for endotoxin assay“, Appl. Microbiol., Bd. 26, Nr. 1, S. 43–48, Juli 1973.
[36] W. K. Bleeker, E. M. Kannegieter, J. C. Bakker, und J. A. Loos, „Endotoxin in blood products:
correlation between the Limulus assay and the rabbit pyrogen test“, Prog. Clin. Biol. Res.,
Bd. 189, S. 293–303, 1985.
[37] C.-Y. Park, S.-H. Jung, J.-P. Bak, S.-S. Lee, und D.-K. Rhee, „Comparison of the rabbit pyrogen
test and Limulus amoebocyte lysate (LAL) assay for endotoxin in hepatitis B vaccines and
the effect of aluminum hydroxide“, Biol. J. Int. Assoc. Biol. Stand., Bd. 33, Nr. 3, S. 145–
151, Sep. 2005.
[38] The Japanese pharmacopoeia: official from April 1, 2011. Tokyo: Pharmaceutical and
Medical Device Regulatory Science Society of Japan, 2012.
[39] United States Pharmacopeial Convention, The United States Pharmacopeia 2011: USP 35 ;
The national formulary : NF 30. Rockville, MD: United States Pharmacopeial Convention,
2011.
99
[40] European Pharmacopoeia, 8th edition 2013, English Subscription to Main volume +
Supplement 1 + Supplement 2. Stuttgart: Deutscher Apotheker Verlag, 2013.
[41] J. L. Ding und B. Ho, „Endotoxin detection--from limulus amebocyte lysate to recombinant
factor C“, Subcell. Biochem., Bd. 53, S. 187–208, 2010.
[42] H. Grallert, S. Leopoldseder, M. Schuett, P. Kurze, und B. Buchberger, „EndoLISA: a novel
and reliable method for endotoxin detection“, Nat Meth, Bd. 8, Nr. 10, Okt. 2011.
[43] European Directorate for the Quality of MedicinesEDQM, „Bacterial endotoxins Ph. Eur.
policy for substances for pharmaceutical use“. Approved by the Ph. Eur. Commission at its
149th Session, 2014.
[44] J. Chen und A. Vinther, „Low Endotoxin Recovery (LER) in Common Biologics Products“,
gehalten auf der Annual PDA meeting, Orlando, 2013.
[45] R. Mello, „LER: An FDA Reviewer´s Perspective“, gehalten auf der PMF Bacterial Endotoxin
Summit, Philadelphia, 16-Mai-2014.
[46] V. Liebers, M. Raulf-Heimsoth, und T. Brüning, „Health effects due to endotoxin inhalation
(review)“, Arch. Toxicol., Bd. 82, Nr. 4, S. 203–210, Apr. 2008.
[47] D. Petsch, W. D. Deckwer, und F. B. Anspach, „Proteinase K digestion of proteins improves
detection of bacterial endotoxins by the Limulus amebocyte lysate assay: application for
endotoxin removal from cationic proteins“, Anal. Biochem., Bd. 259, Nr. 1, S. 42–47, Mai
1998.
[48] A. Gnauck, R. G. Lentle, und M. C. Kruger, „The Limulus Amebocyte Lysate assay may be
unsuitable for detecting endotoxin in blood of healthy female subjects“, J. Immunol.
Methods, Bd. 416, S. 146–156, Jan. 2015.
[49] K. Z. McCullough und Parenteral Drug Association, The bacterial endotoxins test: a
practical approach. Bethesda, MD, USA; River Grove, IL: PDA ; DHI Pub., 2011.
[50] J. C. Hurley, „Endotoxemia: methods of detection and clinical correlates“, Clin. Microbiol.
Rev., Bd. 8, Nr. 2, S. 268–292, Apr. 1995.
[51] E. Ribi, R. L. Anacker, R. Brown, W. T. Haskins, B. Malmgren, K. C. Milner, und J. A. Rudbach,
„Reaction of endotoxin and surfactants. I. Physical and biological properties of endotoxin
treated with sodium deoxycholate“, J. Bacteriol., Bd. 92, Nr. 5, S. 1493–1509, Nov. 1966.
[52] J. A. Rudbach und K. C. Milner, „Reaction of endotoxin and surfactants. III. Effect of sodium
lauryl sulfate on the structure and pyrogenicity of endotoxin“, Can. J. Microbiol., Bd. 14,
Nr. 11, S. 1173–1178, Nov. 1968.
[53] D. K. Milton, H. A. Feldman, D. S. Neuberg, R. J. Bruckner, und I. A. Greaves, „Environmental
endotoxin measurement: The Kinetic Limulus Assay with Resistant-parallel-line
Estimation“, Environ. Res., Bd. 57, Nr. 2, S. 212–230, Apr. 1992.
100
[54] P. Hughes, C. Thomas, K. Suvarna, B. Chi, R. Candau-Chacon, C. Gomez-Broughton, und L.
R. Narasimhan, „Low Endotoxin Recovery: An FDA Perspective“, BioPharma Asia, Bd. 4, Nr.
2, S. 14–25, 04 2015.
[55] K. L. Williams, „Endotoxin Test Concerns of Biologics“, Am. Pharm. Rev., Bd. 16, Nr. 6
Endotoxin Detection Supplement, S. 4–9, 2013.
[56] A. L. Daugherty und R. J. Mrsny, „Formulation and delivery issues for monoclonal antibody
therapeutics“, Adv. Drug Deliv. Rev., Bd. 58, Nr. 5–6, S. 686–706, Aug. 2006.
[57] „Guidance for Industry Pyrogen and Endotoxins Testing: Questions and Answers“. Food
and Drug Administration, Juni-2012.
[58] M. Gradzielski, „Kinetics of morphological changes in surfactant systems“, Curr. Opin.
Colloid Interface Sci., Bd. 8, Nr. 4–5, S. 337–345, Nov. 2003.
[59] M. Tsuchiya, „Possible Mechanism of Low Endotoxin Recovery“, Am. Pharm. Rev., Bd. 17.
[60] D. Petsch und F. B. Anspach, „Endotoxin removal from protein solutions“, J. Biotechnol.,
Bd. 76, Nr. 2–3, S. 97–119, Jan. 2000.
[61] S. Snyder, D. Kim, und T. J. McIntosh, „Lipopolysaccharide Bilayer Structure: Effect of
Chemotype, Core Mutations, Divalent Cations, and Temperature †“, Biochemistry (Mosc.),
Bd. 38, Nr. 33, S. 10758–10767, Aug. 1999.
[62] P. Garidel, M. Rappolt, A. B. Schromm, J. Howe, K. Lohner, J. Andrä, M. H. J. Koch, und K.
Brandenburg, „Divalent cations affect chain mobility and aggregate structure of
lipopolysaccharide from Salmonella minnesota reflected in a decrease of its biological
activity“, Biochim. Biophys. Acta BBA - Biomembr., Bd. 1715, Nr. 2, S. 122–131, Sep. 2005.
[63] A. F. Holleman, E. Wiberg, N. Wiberg, und Holleman-Wiberg, Lehrbuch der anorganischen
Chemie, 101., Und stark Erw. Aufl. Berlin [u.a.]: de Gruyter, 1995.
[64] G. Jander, E. Blasius, J. Strähle, E. Schweda, R. Rossi, und Jander-Blasius, Lehrbuch der
analytischen und präparativen anorganischen Chemie, 16., Überarb. Aufl. Stuttgart: Hirzel,
2006.
[65] A. B. Schromm, K. Brandenburg, E. T. Rietschel, H.-D. Flad, S. F. Carroll, und U. Seydel,
„Lipopolysaccharide-binding protein mediates CD14-independent intercalation of
lipopolysaccharide into phospholipid membranes“, FEBS Lett., Bd. 399, Nr. 3, S. 267–271,
Dez. 1996.
[66] M. Mueller, B. Lindner, S. Kusumoto, K. Fukase, A. B. Schromm, und U. Seydel, „Aggregates
Are the Biologically Active Units of Endotoxin“, J. Biol. Chem., Bd. 279, Nr. 25, S. 26307–
26313, Juni 2004.
[67] D. F. Evans und H. Wennerström, The colloidal domain: where physics, chemistry, biology,
and technology meet, 2nd ed. New York: Wiley-VCH, 1999.
101
[68] N. S. Tan, M. L. Ng, Y. H. Yau, P. K. Chong, B. Ho, und J. L. Ding, „Definition of endotoxin
binding sites in horseshoe crab factor C recombinant sushi proteins and neutralization of
endotoxin by sushi peptides“, FASEB J. Off. Publ. Fed. Am. Soc. Exp. Biol., Bd. 14, Nr. 12, S.
1801–1813, Sep. 2000.
[69] K. Takayama, Z. Din, P. Mukerjee, P. Cooke, und T. Kirkland, „Physicochemical properties
of the lipopolysaccharide unit that activates B lymphocytes.“, J. Biol. Chem., Bd. 265, Nr.
23, S. 14023–14029, Aug. 1990.
[70] P. Garidel, Y. Kaconis, L. Heinbockel, M. Wulf, S. Gerber, A. Munk, V. Vill, und K.
Brandenburg, „Self-Organisation, Thermotropic and Lyotropic Properties of Glycolipids
Related to their Biological Implications“, Open Biochem. J., Bd. 9, Nr. 1, S. 49–72, Sep. 2015.
[71] P. W. Atkins und J. De Paula, Atkins’ Physical chemistry, 9th ed. Oxford ; New York: Oxford
University Press, 2010.
[72] C. W. Twohy, A. P. Duran, und T. E. Munson, „Endotoxin contamination of parenteral drugs
and radiopharmaceuticals as determined by the limulus amebocyte lysate method“, J.
Parenter. Sci. Technol. Publ. Parenter. Drug Assoc., Bd. 38, Nr. 5, S. 190–201, Okt. 1984.
[73] K. Z. McCullough und C. Weidner-Loeven, „Variability in the LAL test: comparison of three
kinetic methods for the testing of pharmaceutical products“, J. Parenter. Sci. Technol. Publ.
Parenter. Drug Assoc., Bd. 46, Nr. 3, S. 69–72, Juni 1992.
[74] M. Kucki, C. Cavelius, und A. Kraegeloh, „Interference of silica nanoparticles with the
traditional Limulus amebocyte lysate gel clot assay“, Innate Immun., Bd. 20, Nr. 3, S. 327–
336, Apr. 2014.
[75] T. Engel und P. Reid, Physikalische Chemie, Bafög-Ausg. München: Pearson Studium, 2009.
[76] H. D. Hochstein, D. F. Mills, A. S. Outschoorn, und S. C. Rastogi, „The processing and
collaborative assay of a reference endotoxin“, J. Biol. Stand., Bd. 11, Nr. 4, S. 251–260, Okt.
1983.
[77] S. E. Greisman und R. B. Hornick, „Comparative pyrogenic reactivity of rabbit and man to
bacterial endotoxin“, Proc. Soc. Exp. Biol. Med. Soc. Exp. Biol. Med. N. Y. N, Bd. 131, Nr. 4,
S. 1154–1158, Sep. 1969.
[78] K. J. Sweadner, M. Forte, und L. L. Nelsen, „Filtration removal of endotoxin (pyrogens) in
solution in different states of aggregation“, Appl. Environ. Microbiol., Bd. 34, Nr. 4, S. 382–
385, Okt. 1977.
[79] T. T. Evans-Strickfaden, K. H. Oshima, A. K. Highsmith, und E. W. Ades, „Endotoxin removal
using 6,000 molecular weight cut-off polyacrylonitrile (PAN) and polysulfone (PS) hollow
fiber ultrafilters“, PDA J. Pharm. Sci. Technol. PDA, Bd. 50, Nr. 3, S. 154–157, Juni 1996.
102
[80] C. Jeworrek, F. Evers, J. Howe, K. Brandenburg, M. Tolan, und R. Winter, „Effects of Specific
versus Nonspecific Ionic Interactions on the Structure and Lateral Organization of
Lipopolysaccharides“, Biophys. J., Bd. 100, Nr. 9, S. 2169–2177, Mai 2011.
[81] N. Kučerka, E. Papp-Szabo, M.-P. Nieh, T. A. Harroun, S. R. Schooling, J. Pencer, E. A.
Nicholson, T. J. Beveridge, und J. Katsaras, „Effect of Cations on the Structure of Bilayers
Formed by Lipopolysaccharides Isolated from Pseudomonas aeruginosa PAO1“, J. Phys.
Chem. B, Bd. 112, Nr. 27, S. 8057–8062, Juli 2008.
[82] M. Herrmann, E. Schneck, T. Gutsmann, K. Brandenburg, und M. Tanaka, „Bacterial
lipopolysaccharides form physically cross-linked, two-dimensional gels in the presence of
divalent cations“, Soft Matter, Bd. 11, Nr. 30, S. 6037–6044, 2015.
[83] M. Vaara, „Agents that increase the permeability of the outer membrane.“, Microbiol.
Rev., Bd. 56, Nr. 3, S. 395–411, Sep. 1992.
[84] H. Nikaido, „Molecular Basis of Bacterial Outer Membrane Permeability Revisited“,
Microbiol. Mol. Biol. Rev., Bd. 67, Nr. 4, S. 593–656, Dez. 2003.
[85] L. A. Clifton, M. W. A. Skoda, A. P. Le Brun, F. Ciesielski, I. Kuzmenko, S. A. Holt, und J. H.
Lakey, „Effect of Divalent Cation Removal on the Structure of Gram-Negative Bacterial
Outer Membrane Models“, Langmuir, Bd. 31, Nr. 1, S. 404–412, Jan. 2015.
[86] M. Walser, „ION ASSOCIATION. V. DISSOCIATION CONSTANTS FOR COMPLEXES OF
CITRATE WITH SODIUM, POTASSIUM, CALCIUM, AND MAGNESIUM IONS 1“, J. Phys. Chem.,
Bd. 65, Nr. 1, S. 159–161, Jan. 1961.
[87] S. Gła̧b, M. Maj-Zurawska, P. łukomski, A. Hulanicki, und A. Lewenstam, „Ion-selective
electrode control based on coulometrically determined stability constants of biologically
important calcium and magnesium complexes“, Anal. Chim. Acta, Bd. 273, Nr. 1–2, S. 493–
497, Feb. 1993.
[88] A. K. Covington und E. Y. Danish, „Measurement of Magnesium Stability Constants of
Biologically Relevant Ligands by Simultaneous Use of pH and Ion-Selective Electrodes“, J.
Solut. Chem., Bd. 38, Nr. 11, S. 1449–1462, Nov. 2009.
[89] Y. Fujita, T. Tokunaga, und H. Kataoka, „Saline and buffers minimize the action of
interfering factors in the bacterial endotoxins test“, Anal. Biochem., Bd. 409, Nr. 1, S. 46–
53, Feb. 2011.
[90] J. Levin und F. B. Bang, „Clottable protein in Limulus; its localization and kinetics of its
coagulation by endotoxin“, Thromb. Diath. Haemorrh., Bd. 19, Nr. 1, S. 186–197, März
1968.
103
[91] J. A. Rudbach, F. I. Akiya, R. J. Elin, H. D. Hochstein, M. K. Luoma, E. C. Milner, K. C. Milner,
und K. R. Thomas, „Preparation and properties of a national reference endotoxin“, J. Clin.
Microbiol., Bd. 3, Nr. 1, S. 21–25, Jan. 1976.
[92] I. Mattsby-Baltzer, K. Lindgren, B. Lindholm, und L. Edebo, „Endotoxin shedding by
enterobacteria: free and cell-bound endotoxin differ in Limulus activity“, Infect. Immun.,
Bd. 59, Nr. 2, S. 689–695, Feb. 1991.
[93] K. Bowers, und T. Lynn, „Creation of an in-house naturally occurring endotoxin preparation
for use in endotoxin spiking studies and LAL sample hold time analysis“, Am. Pharm. Rev.,
Nr. 14, S. 92–97, 2011.
[94] J. S. Bolden, M. E. Claerbout, M. K. Miner, M. A. Murphy, K. R. Smith, und R. E. Warburton,
„Evidence Against a Bacterial Endotoxin Masking Effect in Biologic Drug Products by
Limulus Amebocyte Lysate Detection“, PDA J. Pharm. Sci. Technol., Bd. 68, Nr. 5, S. 472–
477, Sep. 2014.
[95] A. Yan, Z. Guan, und C. R. H. Raetz, „An Undecaprenyl Phosphate-Aminoarabinose Flippase
Required for Polymyxin Resistance in Escherichia coli“, J. Biol. Chem., Bd. 282, Nr. 49, S.
36077–36089, Dez. 2007.
[96] Y. Rosenfeld und Y. Shai, „Lipopolysaccharide (Endotoxin)-host defense antibacterial
peptides interactions: Role in bacterial resistance and prevention of sepsis“, Biochim.
Biophys. Acta BBA - Biomembr., Bd. 1758, Nr. 9, S. 1513–1522, Sep. 2006.
[97] M. Caroff, D. Karibian, J.-M. Cavaillon, und N. Haeffner-Cavaillon, „Structural and
functional analyses of bacterial lipopolysaccharides“, Microbes Infect., Bd. 4, Nr. 9, S. 915–
926, Juli 2002.
[98] D. R. Dixon und R. P. Darveau, „Lipopolysaccharide heterogeneity: innate host responses
to bacterial modification of lipid a structure“, J. Dent. Res., Bd. 84, Nr. 7, S. 584–595, Juli
2005.
[99] Y. Makimura, Y. Asai, A. Sugiyama, und T. Ogawa, „Chemical structure and
immunobiological activity of lipid A from Serratia marcescens LPS“, J. Med. Microbiol., Bd.
56, Nr. 11, S. 1440–1446, Nov. 2007.
[100] Z. Sidorczyk, U. Zahringer, und E. T. Rietschel, „Chemical structure of the lipid A component
of the lipopolysaccharide from a Proteus mirabilis Re-mutant“, Eur. J. Biochem., Bd. 137,
Nr. 1–2, S. 15–22, Dez. 1983.
[101] A. D. Vinion-Dubiel und J. B. Goldberg, „Review: Lipopolysaccharide of Burkholderia
cepacia complex“, J. Endotoxin Res., Bd. 9, Nr. 4, S. 201–213, Aug. 2003.
104
[102] A. Silipo, „Complete structural characterization of the lipid A fraction of a clinical strain of
B. cepacia genomovar I lipopolysaccharide“, Glycobiology, Bd. 15, Nr. 5, S. 561–570, Dez.
2004.
[103] E. A. Groisman, J. Kayser, und F. C. Soncini, „Regulation of polymyxin resistance and
adaptation to low-Mg2+ environments“, J. Bacteriol., Bd. 179, Nr. 22, S. 7040–7045, Nov.
1997.
[104] T. J. Beveridge, „Structures of gram-negative cell walls and their derived membrane
vesicles“, J. Bacteriol., Bd. 181, Nr. 16, S. 4725–4733, Aug. 1999.
[105] C. R. Casella und T. C. Mitchell, „Putting endotoxin to work for us: Monophosphoryl lipid A
as a safe and effective vaccine adjuvant“, Cell. Mol. Life Sci., Bd. 65, Nr. 20, S. 3231–3240,
Okt. 2008.
[106] C. J. van Oss, „A review of “ Intermolecular and Surface Forces, second edition. Jacob N.
Israelachvili. Academic Press, London, 1991. Pp. xxi + 450; hardbound, $49.95.“, J. Dispers.
Sci. Technol., Bd. 13, Nr. 6, S. 718–719, Dez. 1992.
[107] E. T. Rietschel, T. Kirikae, F. U. Schade, A. J. Ulmer, O. Holst, H. Brade, G. Schmidt, U.
Mamat, H.-D. Grimmecke, S. Kusumoto, und U. Zähringer, „The chemical structure of
bacterial endotoxin in relation to bioactivity“, Immunobiology, Bd. 187, Nr. 3–5, S. 169–
190, Apr. 1993.
[108] B. S. Park, D. H. Song, H. M. Kim, B.-S. Choi, H. Lee, und J.-O. Lee, „The structural basis of
lipopolysaccharide recognition by the TLR4–MD-2 complex“, Nature, Bd. 458, Nr. 7242, S.
1191–1195, Apr. 2009.
[109] U. Seydel, L. Hawkins, A. B. Schromm, H. Heine, O. Scheel, M. H. J. Koch, und K.
Brandenburg, „The generalized endotoxic principle“, Eur. J. Immunol., Bd. 33, Nr. 6, S.
1586–1592, Juni 2003.
[110] K. Brandenburg, H. Mayer, M. H. J. Koch, J. Weckesser, E. T. Rietschel, und U. Seydel,
„Influence of the supramolecular structure of free lipid A on its biological activity“, Eur. J.
Biochem., Bd. 218, Nr. 2, S. 555–563, Dez. 1993.
[111] R. Zana, „Aqueous surfactant-alcohol systems: A review“, Adv. Colloid Interface Sci., Bd.
57, S. 1–64, Mai 1995.
[112] M. Tomšič, M. Bešter-Rogač, A. Jamnik, W. Kunz, D. Touraud, A. Bergmann, und O. Glatter,
„Ternary systems of nonionic surfactant Brij 35, water and various simple alcohols:
Structural investigations by small-angle X-ray scattering and dynamic light scattering“, J.
Colloid Interface Sci., Bd. 294, Nr. 1, S. 194–211, Feb. 2006.
[113] D. Langevin, „Micelles and microemulsions“, in Complex Fluids, Bd. 415, L. Garrido, Hrsg.
Berlin, Heidelberg: Springer Berlin Heidelberg, 1993, S. 327–349.
105
[114] A. C. John und A. K. Rakshit, „Effects of mixed alkanols as cosurfactants on single phase
microemulsion properties“, Colloids Surf. Physicochem. Eng. Asp., Bd. 95, Nr. 2–3, S. 201–
210, Feb. 1995.
[115] M. Ruiz-Peña, R. Oropesa-Nuñez, T. Pons, S. R. W. Louro, und A. Pérez-Gramatges,
„Physico-chemical studies of molecular interactions between non-ionic surfactants and
bovine serum albumin“, Colloids Surf. B Biointerfaces, Bd. 75, Nr. 1, S. 282–289, Jan. 2010.
[116] P. Garidel, C. Hoffmann, und A. Blume, „A thermodynamic analysis of the binding
interaction between polysorbate 20 and 80 with human serum albumins and
immunoglobulins: A contribution to understand colloidal protein stabilisation“, Biophys.
Chem., Bd. 143, Nr. 1–2, S. 70–78, Juli 2009.
[117] „Polysorbate 20“, https://commons.wikimedia.org/w/index.php?curid=3649388. 15-Mai-
2016.
[118] „Polysorbate 80“, https://commons.wikimedia.org/w/index.php?curid=3649288. 15-Mai-
2016.
[119] „Triton X-100“, https://commons.wikimedia.org/w/index.php?curid=2264368. 15-Mai-
2016.
[120] U. Sivars und F. Tjerneld, „Mechanisms of phase behaviour and protein partitioning in
detergent/polymer aqueous two-phase systems for purification of integral membrane
proteins11This work was carried out in the Swedish Center for Bioseparation.“, Biochim.
Biophys. Acta BBA - Gen. Subj., Bd. 1474, Nr. 2, S. 133–146, Apr. 2000.
[121] B. Geetha und A. B. Mandal, „The shape, size, aggregation, hydration, correlation times,
and thermodynamic studies on macromonomer micelles“, J. Chem. Phys., Bd. 105, Nr. 21,
S. 9649, 1996.
[122] D. Linke, „Chapter 34 Detergents“, in Methods in Enzymology, Bd. 463, Elsevier, 2009, S.
603–617.
[123] X. Cui, Y. Jiang, C. Yang, X. Lu, H. Chen, S. Mao, M. Liu, H. Yuan, P. Luo, und Y. Du,
„Mechanism of the Mixed Surfactant Micelle Formation“, J. Phys. Chem. B, Bd. 114, Nr. 23,
S. 7808–7816, Juni 2010.
[124] A. D. Nielsen, K. Borch, und P. Westh, „Thermochemistry of the specific binding of C12
surfactants to bovine serum albumin“, Biochim. Biophys. Acta BBA - Protein Struct. Mol.
Enzymol., Bd. 1479, Nr. 1–2, S. 321–331, Juni 2000.
[125] P. Dubin, J. Principi, B. . Smith, und M. Fallon, „Influence of ionic strength and composition
on the size of mixed micelles of sodium dodecyl sulfate and Triton X-100“, J. Colloid
Interface Sci., Bd. 127, Nr. 2, S. 558–565, Feb. 1989.
106
[126] M. Sammalkorpi, M. Karttunen, und M. Haataja, „Ionic Surfactant Aggregates in Saline
Solutions: Sodium Dodecyl Sulfate (SDS) in the Presence of Excess Sodium Chloride (NaCl)
or Calcium Chloride (CaCl 2 )“, J. Phys. Chem. B, Bd. 113, Nr. 17, S. 5863–5870, Apr. 2009.
[127] K. Z. McCullough und Parenteral Drug Association, The bacterial endotoxins test: a
practical approach. Bethesda, MD, USA; River Grove, IL: PDA ; DHI Pub., 2011.
[128] D. Verthelyi und V. Wang, „Trace Levels of Innate Immune Response Modulating Impurities
(IIRMIs) Synergize to Break Tolerance to Therapeutic Proteins“, PLoS ONE, Bd. 5, Nr. 12, S.
e15252, Dez. 2010.
[129] L. A. Haile, M. Puig, L. Kelley-Baker, und D. Verthelyi, „Detection of Innate Immune
Response Modulating Impurities in Therapeutic Proteins“, PLOS ONE, Bd. 10, Nr. 4, S.
e0125078, Apr. 2015.
107
5.4 List of publications
5.4.1 Selected poster presentations
Evaluation of two new recombinant Factor C based assays as alternatives for Limulus blood based
endotoxin detection methods, European Society for Alternatives to Animal Testing, Linz, Austria,
2013
Low Endotoxin Recovery in Limulus Based Detection Systems, Dechema Gesellschaft für
Chemische Technik und Biotechnologie, Irsee, Germany, 2013
Case study - Low Endotoxin Recovery in bio-pharmaceuticals: Comparison of Natural occurring
Endotoxins (NOE) and commercial standards Annual meeting, Parenteral Drug Association, San
Antonio, USA, 2014
Endotoxin Contamination in Biopharmaceuticals: False Negative Results Induced by Endotoxin
Masking, Bioprocessing Summit, Boston, USA, 2014
Endotoxin Testing – A Gamble on the Test System, Annual meeting, Parenteral Drug Association,
Las Vegas, USA, 2015
Challenges of Endotoxin Detection in Biologics Drug Products, Bioprocessing Summit, Boston,
USA, 2015
5.4.2 Oral presentations
An advanced Endotoxin Assay: Insights and Strategies for overcoming Low Endotoxin Recovery in
complex formulations, Parenteral Drug Association, Tokio, Japan, 2013
New Challenges of Endotoxin Detection in modern Pharmaceuticals, International Endotoxin and
Innate Immunity Society, Salt Lake City, USA, 2014
Endotoxin Contamination in BioPharmaceuticals: Overcoming False Negative Results Induced by
Endotoxin Masking, Bioprocessing Summit, Boston, USA, 2014
Endotoxin masking & de-masking, Food and Drug Administration, Bethesda, USA, 2014
Heterogeneity of Potential Endotoxin Contaminations in Parenteral Drugs, Parenteral Drug
Association, Pharmaceutical Microbiology, Berlin, Germany, 2015
Understanding and Overcoming LER, Lonza Endotoxin Summit, Annapolis, USA, 2015
Understanding the Principles of Endotoxin Masking and Demasking, Bacterial Endotoxin Summit,
Iselin, USA, 2015
108
Endotoxin Masking /Low Endotoxin Recovery Update, Webinar, European Compliance Academy,
Heidelberg, Germany, 2015
Overcoming Endotoxin Masking in a Drug Product, PharmaLab, Düsseldorf, Germany, 2015
Masking and Demasking of Endotoxins in Common Biological Product Matrices, Annual Meeting,
Parenteral Drug Association, San Antonio, USA, 2016
Detectability of Endotoxin Contaminations in Biologicals, Protein & Antibody Engineering and
Development Summit, Shanghai, China, 216
Understanding and Overcoming LER II, Lonza Endotoxin Summit, Annapolis, USA, 2016
5.4.3 Published articles
J. Reich, K. Heed, H. Grallert, Detection of naturally occurring bacterial endotoxins in water
samples, European Pharmaceutical Review magazine, Issue 6, 2014
Z. Hu, T. Murakami, K. Suzuki, H. Tamura, J. Reich, K. Kuwahara-Arai, T. Iba, und I. Nagaoka,
„Antimicrobial cathelicidin peptide LL-37 inhibits the pyroptosis of macrophages and improves
the survival of polybacterial septic mice“, Int. Immunol., 28, 5, 2016.
L. Wimbish, J. Reich, Better understanding LER and how to detect endotoxins in medicinal
products, Pharmaceutical Processing, Vol 31, No. 3, 2016
J. Reich, P. Lang, H. Grallert, H. Motschmann, Masking of Endotoxin in Surfactant Samples: Effects
on Limulus-based detection systems, Biologicals, 2016 (in press)
H. Tamura, J. Reich, I. Nagaoka, Bacterial endotoxin assays relevant to host defense peptides,
submitted to Juntendo Medical Journal, 62, 2, 2016
5.4.4 Intellectual Property
J. Reich, H. Grallert, Unmasking endotoxins in solution, EP20140172151, 2014
109
5.4.5 Declaration
The studies presented in chapter 3.1 “Masking of endotoxin in surfactant samples: Effects on
Limulus-based detection systems” led to a publication which is was already submitted to the
Journal of Biologicals in 2015 (J. Reich, P. Lang, H. Grallert, H. Motschmann, Masking of Endotoxin
in Surfactant Samples: Effects on Limulus-based detection systems). The article has been
accepted on 26th of April 2016 and is currently in press.
The findings described in chapter 3.4 “Endotoxin demasking” are included in a patent application
in 2014 to protect this unique approach. Therefore, parts of this chapter (3.4) were already
published in 2015 (J. Reich, H. Grallert, Unmasking endotoxins in solution, EP20140172151, 2014).
110
5.5 Curriculum Vitae
Johannes Reich
Address: An der Bärenmühle 6
82362 Weilheim, Germany
E-mail: [email protected]
Employment History and Experience
09/2016 – Microcoat Biotechnologie GmbH, Bernried, Germany General Manager - endotoxin test service
02/2005 – 12/2008 Profos AG, Regensburg, Germany Product Manager - drugs of abuse (DoA) testing product range (whilst concluding final elements of University degree)
08/2004 – 01/2005 Profos AG, Regensburg, Germany Marketing Assistant - all technology/product categories. Industrial year during University degree course.
01/2002 – 08/2002 Raiffeisen-BayWa Waren GmbH, Lobsing, Germany Management Assistant - wholesale outlet
Education and Training
10/2012 – 08/2016 Universität Regensburg, Regensburg, Germany Hyglos GmbH, Bernried, Germany Doctoral Research Study - Low Endotoxin Recovery
10/2010 – 09/2012 Universität Regensburg, Regensburg, Germany Institute for Separative Chemistry, Marcoule, France Master of Science (COSOM)
10/2007 – 09/2010 Universität Regensburg, Regensburg, Germany Bachelor of Science (Chemistry)
09/2003 – 07/2007 University of Applied Sciences, Regensburg, Germany Diploma (Business Administration)
09/2002 – 08/2003 Upper vocational school, BOS, Regensburg, Germany Advanced technical college entrance qualification
03/2001 – 12/2001 Bundeswehr Pionierlehrbatallion, Ingolstadt, Germany Military service -Technical training as mobile crane operator
09/1998 – 02/2001 Raiffeisen-BayWa Waren GmbH, Lobsing, Germany Professional training / work experience - Training in commercial aspects of wholesale and export trades.
111
5.6 Eidesstattliche Erklärung Ich erkläre hiermit an Eides statt, dass ich die vorliegende Arbeit ohne unzulässige Hilfe Dritter
und ohne Benutzung anderer als der angegebenen Hilfsmittel angefertigt habe; die aus anderen
Quellen direkt oder indirekt übernommenen Daten und Konzepte sind unter Angabe des
Literaturzitats gekennzeichnet.
Weitere Personen waren an der inhaltlich-materiellen Herstellung der vorliegenden Arbeit nicht
beteiligt. Insbesondere habe ich hierfür nicht die entgeltliche Hilfe eines Promotionsberaters oder
anderer Personen in Anspruch genommen. Niemand hat von mir weder unmittelbar noch
mittelbar geldwerte Leistungen für Arbeiten erhalten, die im Zusammenhang mit dem Inhalt der
vorgelegten Dissertation stehen.
Die Arbeit wurde bisher weder im In- noch im Ausland in gleicher oder ähnlicher Form einer
anderen Prüfungsbehörde vorgelegt.
Regensburg, den 05.07.2016
______________________________
Johannes Reich
112
113
6 Annex
114
115
116
117
118
119
120
Top Related